<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD with MathML3 v1.1d2 20140930//EN" "JATS-journalpublishing1-mathml3.dtd">
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" dtd-version="1.1d2" xml:lang="en">
  <front>
    <journal-meta>
      <journal-id journal-id-type="nlm-ta">CJIF</journal-id>
      <journal-id journal-id-type="publisher-id">ICCK</journal-id>
      <journal-title-group>
        <journal-title>Chinese Journal of Information Fusion</journal-title>
      </journal-title-group>
      <issn pub-type="ppub" publication-format="print">2998-3363</issn>
      <issn pub-type="epub" publication-format="electronic">2998-3371</issn>
      <publisher>
        <publisher-name>Institute of Central Computation and Knowledge Inc</publisher-name>
        <publisher-loc>522 W RIVERSIDE AVE STE N, SPOKANE, WA, 99201-0508, UNITED STATES</publisher-loc>
      </publisher>
    </journal-meta>
    <article-meta>
      <article-id pub-id-type="doi">10.62762/CJIF.2025.695812</article-id>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Research Article</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>An Improved YOLOv8-Based Detection Model for Multi-Scale Sea Ice in Satellite Imagery</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">https://orcid.org/0009-0002-4342-9280</contrib-id>
          <name>
            <surname>Liu</surname>
            <given-names>Yang</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">https://orcid.org/0009-0005-8845-5468</contrib-id>
          <name>
            <surname>Guo</surname>
            <given-names>Qiang</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Dong</surname>
            <given-names>Chengguo</given-names>
          </name>
          <xref ref-type="aff" rid="aff2">2</xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Luo</surname>
            <given-names>Yiping</given-names>
          </name>
          <xref ref-type="aff" rid="aff3">3</xref>
        </contrib>
        <aff id="aff1"><label>1</label>School of Computer and Control Engineering, Yantai University, Yantai 264005, China</aff>
        <aff id="aff2"><label>2</label>School of Architectural Engineering, Weifang University of Science and Technology, Weifang 261000, China</aff>
        <aff id="aff3"><label>3</label>Deep Space Exploration Laboratory, Hefei 230000, China</aff>
      </contrib-group>
      <author-notes>
        <corresp id="cor2">Corresponding Author: Qiang Guo. Email: <email>guoqiang@ytu.edu.cn</email></corresp>
      </author-notes>
      <pub-date date-type="pub" pub-type="epub" publication-format="online">
        <day>29</day>
        <month>3</month>
        <year>2025</year>
      </pub-date>
      <volume>2</volume>
      <issue>1</issue>
      <fpage>79</fpage>
      <lpage>99</lpage>
      <history>
        <date date-type="received">
          <day>05</day>
          <month>3</month>
          <year>2025</year>
        </date>
        <date date-type="accepted">
          <day>23</day>
          <month>3</month>
          <year>2025</year>
        </date>
      </history>
      <permissions>
        <copyright-statement>© 2025 by the Authors. Published by Institute of Central Computation and Knowledge. This is an open access article under the CC BY license (https://creativecommons.org/licenses/by/4.0/).</copyright-statement>
        <copyright-year>2025</copyright-year>
        <copyright-holder>The Authors</copyright-holder>
        <license xlink:href="https://creativecommons.org/licenses/by/4.0/">
        </license>
      </permissions>
      <self-uri xlink:href="https://www.icck.org/article/abs/cjif.2025.695812">This article is available from https://www.icck.org/article/abs/cjif.2025.695812</self-uri>
      <abstract>
        <p>Sea ice detection is of vital importance for maritime navigation. Satellite imagery is a crucial medium for conveying information about sea ice. Currently, most sea ice detection models mainly rely on texture information to identify sea ice in satellite imagery, while ignoring sea ice size information. This research presents an improved YOLOv8-Based detection algorithm for multi-scale sea ice. First, we propose a fusion module based on the attention mechanism and use it to replace the Concat module in the YOLOv8 network structure. Second, we conduct an applicability analysis of the bounding box regression loss function in YOLOv8 and ultimately select Shape-IoU, which is more suitable for sea ice, as the loss function for bounding box regression. Third, we analyze the distribution characteristics of sea ice with different sizes in the NWPU-RESISC45 dataset. Based on these distribution characteristics, the bounding box information predicted by YOLOv8 are converted into evidence vectors for uncertainty quantification. Subsequently, information fusion is achieved by fusing these vectors with the probability of sea ice categories. Compared to YOLOv8 and other detection algorithm, our improved YOLOv8 achieves better detection accuracy on both the NWPU-RESISC45 and the <ext-link xlink:href="https://github.com/LiuYang0911/A-Proprietary-Visible-Light-based-Sea-Ice-Dataset">Landsat-8-derived Sea Ice</ext-link> datasets.</p>
      </abstract>
      <kwd-group kwd-group-type="author" xml:lang="en">
        <kwd>satellite imagery</kwd>
        <kwd>YOLO</kwd>
        <kwd>attention mechanism</kwd>
        <kwd>loss function</kwd>
        <kwd>information fusion</kwd>
        <kwd>evidential reasoning</kwd>
      </kwd-group>
    </article-meta>
  </front>
  <body>
    <sec id="S1">
      <label>1.</label>
      <title>Introduction</title>
      <p id="S1.p1">In recent years, due to the continuous global warming [<xref rid="ref001" ref-type="bibr">1</xref>, <xref rid="ref002" ref-type="bibr">2</xref>], the sea ice in high-latitude regions has been persistently melting [<xref rid="ref003" ref-type="bibr">3</xref>, <xref rid="ref004" ref-type="bibr">4</xref>]. The resulting high-latitude waterways can shorten the sailing distances between major trading powers and are urgently in need of development as future maritime routes [<xref rid="ref005" ref-type="bibr">5</xref>, <xref rid="ref006" ref-type="bibr">6</xref>, <xref rid="ref007" ref-type="bibr">7</xref>]. Specifically, sea ice detection has always been the focus of research in high-latitude seas, which is devoted to accurately locating the positions of sea ice and identifying the scales of sea ice [<xref rid="ref008" ref-type="bibr">8</xref>, <xref rid="ref009" ref-type="bibr">9</xref>].</p>
      <p id="S1.p2">A multitude of technologies are emerging in the domain of real-time object detection. They are extensively adopted in diverse industries, including the identification of suspicious behavior [<xref rid="ref010" ref-type="bibr">10</xref>], the detection of anomalies in medical images [<xref rid="ref011" ref-type="bibr">11</xref>], and fish detection [<xref rid="ref012" ref-type="bibr">12</xref>], and other applications. In recent years, researchers have been concentrating on designing CNN-based object detectors [<xref rid="ref013" ref-type="bibr">13</xref>, <xref rid="ref014" ref-type="bibr">14</xref>, <xref rid="ref015" ref-type="bibr">15</xref>, <xref rid="ref016" ref-type="bibr">16</xref>, <xref rid="ref017" ref-type="bibr">17</xref>, <xref rid="ref018" ref-type="bibr">18</xref>]. Among them, YOLOs achieve accurate classification and positioning of objects with low latency, and they are increasingly gaining popularity [<xref rid="ref019" ref-type="bibr">19</xref>, <xref rid="ref020" ref-type="bibr">20</xref>, <xref rid="ref021" ref-type="bibr">21</xref>, <xref rid="ref022" ref-type="bibr">22</xref>, <xref rid="ref023" ref-type="bibr">23</xref>, <xref rid="ref024" ref-type="bibr">24</xref>, <xref rid="ref025" ref-type="bibr">25</xref>, <xref rid="ref026" ref-type="bibr">26</xref>, <xref rid="ref027" ref-type="bibr">27</xref>].</p>
      <p id="S1.p3">Furthermore, for an extended period, considerable efforts have been directed towards obtaining high-quality sea ice satellite remote sensing information and detecting sea ice from a diverse range of satellite remote sensing data [<xref rid="ref028" ref-type="bibr">28</xref>, <xref rid="ref029" ref-type="bibr">29</xref>, <xref rid="ref030" ref-type="bibr">30</xref>, <xref rid="ref031" ref-type="bibr">31</xref>]. Hu et al. [<xref rid="ref032" ref-type="bibr">32</xref>] detected sea ice using GNSS bidirectional radar reflections, where the local linear embedding (LLE) algorithm was employed for sea ice feature extraction. Liu et al. [<xref rid="ref033" ref-type="bibr">33</xref>] proposed a Bayesian method with consideration of geometric characteristics of China France Oceanography Satellite scatterometer(CSCAT) for sea ice detection. The method operationally produced daily polar sea ice mask throughout its mission duration from 2019 to 2022. Jafari et al. [<xref rid="ref034" ref-type="bibr">34</xref>] developed an automated method for iceberg detection and classification in complex sea conditions. Using the RADARSAT Constellation Mission (RCM), they collected seasonal sea ice data from the east coast of Canada.</p>
      <p id="S1.p4">To obtain more abundant spectral information, researchers have explored diverse types of optical remote sensing data [<xref rid="ref035" ref-type="bibr">35</xref>, <xref rid="ref036" ref-type="bibr">36</xref>, <xref rid="ref037" ref-type="bibr">37</xref>, <xref rid="ref038" ref-type="bibr">38</xref>]. Researchers have focused on studying sea ice with visible remote sensing data, as the human eye can intuitively perceive the difference between sea ice and seawater in the visible wavelength band. Advancements in deep convolutional neural networks have achieved automated sea ice detection using visible remote sensing data. Ding et al. [<xref rid="ref039" ref-type="bibr">39</xref>] proposed a detection model based on YOLOv5. They added Squeeze-and-Excitation Networks (SE) [<xref rid="ref040" ref-type="bibr">40</xref>] to backbone of YOLOv5. The SE module computes channel-wise attention weights through global average pooling and multilayer perceptron, which are then applied to recalibrate feature map channels by element-wise multiplication. However, over-dependence on channel attention mechanisms (e.g., SE) inevitably discards spatially fine-grained features in imagery, particularly ice-water interface textures and areal extent variations that are critical for sea ice detection.</p>
      <p id="S1.p5">In this paper, we aim to address these questions precisely and further broaden the application scope of YOLOs. Refining the details of YOLOv8, we aim to enhance its capability in identifying sea ice across a variety of sizes.</p>
      <p id="S1.p6">Our contributions are as follows:</p>
      <p>
        <list list-type="order" id="S1.I1">
          <list-item id="S1.I1.i1">
            <p id="S1.I1.i1.p1">First, we propose a fusion module based on the attention mechanism and use it to replace the Concat module in the YOLOv8 network structure. This module can effectively help YOLOv8 extract the characteristic information of sea ice.</p>
          </list-item>
          <list-item id="S1.I1.i2">
            <p id="S1.I1.i2.p1">Second, we conduct an applicability analysis of the bounding box regression loss function in YOLOv8 and ultimately select Shape-IoU, which is more suitable for sea ice, as the loss function for bounding box regression. YOLOv8 utilizing Shape-IoU [<xref rid="ref041" ref-type="bibr">41</xref>] not only demonstrates superior detection accuracy across all three categories of sea ice, but it also significantly reduces convergence time.</p>
          </list-item>
          <list-item id="S1.I1.i3">
            <p id="S1.I1.i3.p1">Third, we analyze the distribution characteristics of sea ice with different sizes in the NWPU-RESISC45 dataset. Based on these distribution characteristics, the bounding box information predicted by YOLOv8 are converted into evidence vectors for uncertainty quantification. Subsequently, evidence fusion [<xref rid="ref042" ref-type="bibr">42</xref>] is achieved by fusing these vectors with the probability of sea ice categories.</p>
          </list-item>
          <list-item id="S1.I1.i4">
            <p id="S1.I1.i4.p1">Based on Landsat-8 satellite data, we have created a sea ice dataset and made it publicly available on this website: <ext-link xlink:href="https://github.com/LiuYang0911/A-Proprietary-Visible-Light-based-Sea-Ice-Dataset">https://github.com/LiuYang0911/A-Proprietary-Visible-Light-based-Sea-Ice-Dataset</ext-link>.</p>
          </list-item>
          <list-item id="S1.I1.i5">
            <p id="S1.I1.i5.p1">By comparing with current mainstream object detection algorithms, our improved YOLOv8 achieves better detection accuracy and faster convergence speed.</p>
          </list-item>
        </list>
      </p>
      <p id="S1.p8">We are optimistic that the outcomes of our efforts can act as a catalyst for the progress of fellow researchers in this domain.</p>
    </sec>
    <sec id="S2">
      <label>2.</label>
      <title>Related Work</title>
      <sec id="S2.SS1">
        <label>2.1</label>
        <title>Attention Mechanism</title>
        <p id="S2.SS1.p1">Initially, attention mechanisms were utilized in machine translation tasks. This mechanism enables the model to focus on different parts of the input sentence when translating a word, which significantly enhances the translation quality [<xref rid="ref043" ref-type="bibr">43</xref>]. Over the past years, considerable efforts have been devoted to developing attention mechanism modules that are more applicable to the domain of computer vision [<xref rid="ref040" ref-type="bibr">40</xref>, <xref rid="ref044" ref-type="bibr">44</xref>, <xref rid="ref045" ref-type="bibr">45</xref>].</p>
        <p>
          <fig id="F1">
            <label>Figure 1.</label>
            <caption>
              <p>Schematic representation of the network architecture for YOLOv8.</p>
            </caption>
            <graphic xlink:href="fig1YOLOv8.pdf"/>
          </fig>
        </p>
        <p>
          <list list-type="order" id="S2.I1">
            <list-item id="S2.I1.i1">
              <p id="S2.I1.i1.p1">Channel attention mechanism: This type of attention mechanism, which concentrates on the channel dimension of the feature map, aims to enhance significant channel information while suppressing less important data. It accomplishes this by learning weights for each channel, akin to the methodology employed in SENet [<xref rid="ref040" ref-type="bibr">40</xref>]. In SENet [<xref rid="ref040" ref-type="bibr">40</xref>], the input feature map is first compressed along the spatial dimension before calculating weights for each channel. Finally, these weights are applied to multiply with the input feature map to produce the final output.</p>
            </list-item>
            <list-item id="S2.I1.i2">
              <p id="S2.I1.i2.p1">Spatial attention mechanism: In contrast to the channel attention mechanism, the spatial attention mechanism emphasizes the locations of valid information within the feature map, as exemplified by STN [<xref rid="ref044" ref-type="bibr">44</xref>]. STN [<xref rid="ref044" ref-type="bibr">44</xref>] is capable of extracting characteristics from significant regions across various deformation data to produce final prediction results.</p>
            </list-item>
            <list-item id="S2.I1.i3">
              <p id="S2.I1.i3.p1">Hybrid Attention Mechanism: Compared to the aforementioned two attention mechanisms, this particular attention mechanism comprehensively leverages both channel information and spatial information from feature maps, as exemplified by CBAM [<xref rid="ref045" ref-type="bibr">45</xref>]. It sequentially employs the channel attention module followed by the spatial attention module to generate attention weights, ultimately producing the final feature map.</p>
            </list-item>
          </list>
        </p>
      </sec>
      <sec id="S2.SS2">
        <label>2.2</label>
        <title>YOLOs</title>
        <sec id="S2.SS2.SSS1">
          <label>2.2.1</label>
          <title>Modules and Network Architecture</title>
          <p id="S2.SS2.SSS1.p1">In 2016, Joseph Redmon introduced YOLOv1 [<xref rid="ref018" ref-type="bibr">18</xref>], a real-time object detector built upon the deep learning framework Darknet. When compared to other object detectors [<xref rid="ref013" ref-type="bibr">13</xref>, <xref rid="ref014" ref-type="bibr">14</xref>, <xref rid="ref015" ref-type="bibr">15</xref>, <xref rid="ref016" ref-type="bibr">16</xref>, <xref rid="ref017" ref-type="bibr">17</xref>], YOLOs [<xref rid="ref018" ref-type="bibr">18</xref>, <xref rid="ref019" ref-type="bibr">19</xref>, <xref rid="ref020" ref-type="bibr">20</xref>] demonstrate superior detection performance while maintaining high detection speeds.</p>
          <p id="S2.SS2.SSS1.p2">Over the past few years, significant efforts have been dedicated to exploring more efficient modules and network architectures for the YOLO series. YOLOv4 [<xref rid="ref021" ref-type="bibr">21</xref>] and YOLOv5 [<xref rid="ref022" ref-type="bibr">22</xref>] investigated the impact of various activation functions on detection accuracy and speed. It is essential to recognize that YOLOv5 [<xref rid="ref022" ref-type="bibr">22</xref>] has been widely adopted across numerous sectors as a highly effective object detector. Building upon RepVGG, YOLOv6 [<xref rid="ref023" ref-type="bibr">23</xref>] introduced RepBlock to replace the CSPDarknet53 [<xref rid="ref046" ref-type="bibr">46</xref>] architecture used in YOLOv5, which allows the model to better integrate multi-scale features. Furthermore, based on YOLOv5 [<xref rid="ref022" ref-type="bibr">22</xref>], YOLOv7 [<xref rid="ref024" ref-type="bibr">24</xref>] proposed E-ELAN [<xref rid="ref047" ref-type="bibr">47</xref>], which enhances the network's learning capability while preserving the original gradient path.</p>
          <p id="S2.SS2.SSS1.p3">Based on the C3 module of YOLOv5 [<xref rid="ref022" ref-type="bibr">22</xref>], YOLOv8 [<xref rid="ref025" ref-type="bibr">25</xref>] has developed the C2f module, as illustrated in Figure <xref ref-type="fig" rid="F1">1</xref>. This module dynamically adjusts the number of channels according to the model's size, enabling it to flexibly adapt to various scenarios.</p>
          <p id="S2.SS2.SSS1.p4">YOLOv9 [<xref rid="ref026" ref-type="bibr">26</xref>] introduced G-ELEN, a network architecture that integrates the features of CSPNet [<xref rid="ref046" ref-type="bibr">46</xref>] and ELEN [<xref rid="ref047" ref-type="bibr">47</xref>], aiming to enhance detection accuracy while preserving detection speed. Building on the foundation established by YOLOv8 [<xref rid="ref025" ref-type="bibr">25</xref>], YOLOv10 [<xref rid="ref027" ref-type="bibr">27</xref>] presented several improvements, including the use of classification heads with reduced parameters and the incorporation of a partial self-attention module, etc., all designed to further transcend the accuracy-speed trade-offs inherent in YOLO models.</p>
        </sec>
        <sec id="S2.SS2.SSS2">
          <label>2.2.2</label>
          <title>Loss Function Utilized in Bounding Box Regression</title>
          <p id="S2.SS2.SSS2.p1">The object detector based on convolutional neural networks employs a loss function to update the network weights [<xref rid="ref048" ref-type="bibr">48</xref>]. Historically, iterations of the YOLO object detectors have been engaged in an unrelenting pursuit of optimizing the loss function for bounding box regression, aiming to achieve superior performance [<xref rid="ref048" ref-type="bibr">48</xref>, <xref rid="ref049" ref-type="bibr">49</xref>, <xref rid="ref050" ref-type="bibr">50</xref>]. Simultaneously, a variety of loss functions for boundary box regression are continuously being developed and refined [<xref rid="ref041" ref-type="bibr">41</xref>, <xref rid="ref051" ref-type="bibr">51</xref>, <xref rid="ref052" ref-type="bibr">52</xref>, <xref rid="ref053" ref-type="bibr">53</xref>, <xref rid="ref054" ref-type="bibr">54</xref>], thereby enabling the improved YOLOv8-based object detector to be applied across an increasingly diverse range of scenarios.</p>
          <p>
            <fig id="F2">
              <label>Figure 2.</label>
              <caption>
                <p>Schematic representation of the loss function for bounding box regression.</p>
              </caption>
              <graphic xlink:href="fig2IoU.pdf"/>
            </fig>
          </p>
          <p id="S2.SS2.SSS2.p2">The YOLOv8 model employs C-IoU [<xref rid="ref050" ref-type="bibr">50</xref>] as the loss function for bounding box regression, as shown in Figure <xref ref-type="fig" rid="F2">2</xref>. The following presents the mathematical expression for C-IoU [<xref rid="ref050" ref-type="bibr">50</xref>]:</p>
          <p>
            <disp-formula id="S2.E1">
              <mml:math alttext="C{}IoU=IoU-\frac{(x-x_{gt})^{2}+(y-y_{gt})^{2}}{W^{2}+H^{2}}-\alpha\upsilon" display="block">
                <mml:mrow>
                  <mml:mrow>
                    <mml:mi>C</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>I</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>o</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>U</mml:mi>
                  </mml:mrow>
                  <mml:mo>=</mml:mo>
                  <mml:mrow>
                    <mml:mrow>
                      <mml:mi>I</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>o</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>U</mml:mi>
                    </mml:mrow>
                    <mml:mo>−</mml:mo>
                    <mml:mfrac>
                      <mml:mrow>
                        <mml:msup>
                          <mml:mrow>
                            <mml:mo stretchy="false">(</mml:mo>
                            <mml:mrow>
                              <mml:mi>x</mml:mi>
                              <mml:mo>−</mml:mo>
                              <mml:msub>
                                <mml:mi>x</mml:mi>
                                <mml:mrow>
                                  <mml:mi>g</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>t</mml:mi>
                                </mml:mrow>
                              </mml:msub>
                            </mml:mrow>
                            <mml:mo stretchy="false">)</mml:mo>
                          </mml:mrow>
                          <mml:mn>2</mml:mn>
                        </mml:msup>
                        <mml:mo>+</mml:mo>
                        <mml:msup>
                          <mml:mrow>
                            <mml:mo stretchy="false">(</mml:mo>
                            <mml:mrow>
                              <mml:mi>y</mml:mi>
                              <mml:mo>−</mml:mo>
                              <mml:msub>
                                <mml:mi>y</mml:mi>
                                <mml:mrow>
                                  <mml:mi>g</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>t</mml:mi>
                                </mml:mrow>
                              </mml:msub>
                            </mml:mrow>
                            <mml:mo stretchy="false">)</mml:mo>
                          </mml:mrow>
                          <mml:mn>2</mml:mn>
                        </mml:msup>
                      </mml:mrow>
                      <mml:mrow>
                        <mml:msup>
                          <mml:mi>W</mml:mi>
                          <mml:mn>2</mml:mn>
                        </mml:msup>
                        <mml:mo>+</mml:mo>
                        <mml:msup>
                          <mml:mi>H</mml:mi>
                          <mml:mn>2</mml:mn>
                        </mml:msup>
                      </mml:mrow>
                    </mml:mfrac>
                    <mml:mo>−</mml:mo>
                    <mml:mrow>
                      <mml:mi>α</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>υ</mml:mi>
                    </mml:mrow>
                  </mml:mrow>
                </mml:mrow>
              </mml:math>
            </disp-formula>
          </p>
          <p>
            <disp-formula id="S2.E2">
              <mml:math alttext="IoU=\frac{B_{Anchor}\cap B_{Ground~{}Truth}}{B_{Anchor}\cup B_{Ground~{}Truth}}" display="block">
                <mml:mrow>
                  <mml:mrow>
                    <mml:mi>I</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>o</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>U</mml:mi>
                  </mml:mrow>
                  <mml:mo>=</mml:mo>
                  <mml:mfrac>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>B</mml:mi>
                        <mml:mrow>
                          <mml:mi>A</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>n</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>c</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>h</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>r</mml:mi>
                        </mml:mrow>
                      </mml:msub>
                      <mml:mo>∩</mml:mo>
                      <mml:msub>
                        <mml:mi>B</mml:mi>
                        <mml:mrow>
                          <mml:mi>G</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>r</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>u</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>n</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>d</mml:mi>
                          <mml:mo lspace="0.230em">⁢</mml:mo>
                          <mml:mi>T</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>r</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>u</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>t</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>h</mml:mi>
                        </mml:mrow>
                      </mml:msub>
                    </mml:mrow>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>B</mml:mi>
                        <mml:mrow>
                          <mml:mi>A</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>n</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>c</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>h</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>r</mml:mi>
                        </mml:mrow>
                      </mml:msub>
                      <mml:mo>∪</mml:mo>
                      <mml:msub>
                        <mml:mi>B</mml:mi>
                        <mml:mrow>
                          <mml:mi>G</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>r</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>u</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>n</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>d</mml:mi>
                          <mml:mo lspace="0.230em">⁢</mml:mo>
                          <mml:mi>T</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>r</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>u</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>t</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>h</mml:mi>
                        </mml:mrow>
                      </mml:msub>
                    </mml:mrow>
                  </mml:mfrac>
                </mml:mrow>
              </mml:math>
            </disp-formula>
          </p>
          <p>
            <disp-formula id="S2.E3">
              <mml:math alttext="\alpha=\frac{\upsilon}{1-IoU+\upsilon}" display="block">
                <mml:mrow>
                  <mml:mi>α</mml:mi>
                  <mml:mo>=</mml:mo>
                  <mml:mfrac>
                    <mml:mi>υ</mml:mi>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mn>1</mml:mn>
                        <mml:mo>−</mml:mo>
                        <mml:mrow>
                          <mml:mi>I</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>U</mml:mi>
                        </mml:mrow>
                      </mml:mrow>
                      <mml:mo>+</mml:mo>
                      <mml:mi>υ</mml:mi>
                    </mml:mrow>
                  </mml:mfrac>
                </mml:mrow>
              </mml:math>
            </disp-formula>
          </p>
          <p>
            <disp-formula id="S2.E4">
              <mml:math alttext="\upsilon=\frac{4}{\pi^{2}}\ast(\tan^{-1}\frac{w}{h}-\tan^{-1}\frac{w_{gt}}{h_{%&#10;gt}})^{2}" display="block">
                <mml:mrow>
                  <mml:mi>υ</mml:mi>
                  <mml:mo>=</mml:mo>
                  <mml:mrow>
                    <mml:mfrac>
                      <mml:mn>4</mml:mn>
                      <mml:msup>
                        <mml:mi>π</mml:mi>
                        <mml:mn>2</mml:mn>
                      </mml:msup>
                    </mml:mfrac>
                    <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                    <mml:msup>
                      <mml:mrow>
                        <mml:mo stretchy="false">(</mml:mo>
                        <mml:mrow>
                          <mml:mrow>
                            <mml:msup>
                              <mml:mi>tan</mml:mi>
                              <mml:mrow>
                                <mml:mo>−</mml:mo>
                                <mml:mn>1</mml:mn>
                              </mml:mrow>
                            </mml:msup>
                            <mml:mo lspace="0.167em">⁡</mml:mo>
                            <mml:mfrac>
                              <mml:mi>w</mml:mi>
                              <mml:mi>h</mml:mi>
                            </mml:mfrac>
                          </mml:mrow>
                          <mml:mo>−</mml:mo>
                          <mml:mrow>
                            <mml:msup>
                              <mml:mi>tan</mml:mi>
                              <mml:mrow>
                                <mml:mo>−</mml:mo>
                                <mml:mn>1</mml:mn>
                              </mml:mrow>
                            </mml:msup>
                            <mml:mo lspace="0.167em">⁡</mml:mo>
                            <mml:mfrac>
                              <mml:msub>
                                <mml:mi>w</mml:mi>
                                <mml:mrow>
                                  <mml:mi>g</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>t</mml:mi>
                                </mml:mrow>
                              </mml:msub>
                              <mml:msub>
                                <mml:mi>h</mml:mi>
                                <mml:mrow>
                                  <mml:mi>g</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>t</mml:mi>
                                </mml:mrow>
                              </mml:msub>
                            </mml:mfrac>
                          </mml:mrow>
                        </mml:mrow>
                        <mml:mo stretchy="false">)</mml:mo>
                      </mml:mrow>
                      <mml:mn>2</mml:mn>
                    </mml:msup>
                  </mml:mrow>
                </mml:mrow>
              </mml:math>
            </disp-formula>
          </p>
          <p id="S2.SS2.SSS2.p3">From the formulas, it is evident that C-IoU [<xref rid="ref050" ref-type="bibr">50</xref>] takes into account both the position and shape of the bounding box in a comprehensive manner. This allows the model to learn the characteristics of the ground truth box more thoroughly.</p>
        </sec>
      </sec>
    </sec>
    <sec id="S3">
      <label>3.</label>
      <title>Methodology</title>
      <sec id="S3.SS1">
        <label>3.1</label>
        <title>An Attention-Based Fusion Module</title>
        <p id="S3.SS1.p1">We categorize these sea ice instances into three distinct groups, as elaborated in section <xref rid="S4.SS1">4.1</xref>. Although satellite imagery offers relatively high resolution, actual ice conditions can be highly complex. As shown in Figure <xref ref-type="fig" rid="F3">3</xref>, several factors make it challenging for object detection models to accurately identify sea ice of varying scales. These include the wide range of ice floe sizes, irregular shapes, and reduced contrast between ice and seawater caused by melting and accumulation of sea ice.</p>
        <p>
          <fig id="F3">
            <label>Figure 3.</label>
            <caption>
              <p>The two key issues: (a) Numerous small-scale sea ice; (b) Ambiguous demarcation between sea ice and seawater.</p>
            </caption>
            <graphic xlink:href="fig3.pdf"/>
          </fig>
        </p>
        <p>
          <fig id="F4">
            <label>Figure 4.</label>
            <caption>
              <p>Schematic representation of the Attention-Based Fusion Module.</p>
            </caption>
            <graphic xlink:href="fig4.pdf"/>
          </fig>
        </p>
        <p id="S3.SS1.p2">Besides, as the network deepens, the detection model progressively enhances semantic information in feature maps while inevitably sacrificing spatial details, particularly size characteristics crucial for sea ice analysis. When handling this task, YOLOv8 uses the Concat module to combine deep and shallow feature maps. However, we observe that relying on the feature maps after direct stitching is not sufficient for accurate size classification. To address this limitation, we propose an attention-based fusion module that can effectively enhance the spatial detail information in the feature map, so as to be able to accurately distinguish between sea ice size categories, as shown in Figure <xref ref-type="fig" rid="F4">4</xref>.</p>
        <p id="S3.SS1.p3">First, we calculate the channel attention weight <inline-formula><mml:math alttext="M_{C}" display="inline"><mml:msub><mml:mi>M</mml:mi><mml:mi>C</mml:mi></mml:msub></mml:math></inline-formula> of the feature map <inline-formula><mml:math alttext="F_{1}" display="inline"><mml:msub><mml:mi>F</mml:mi><mml:mn>1</mml:mn></mml:msub></mml:math></inline-formula> and multiply it with the feature map <inline-formula><mml:math alttext="F_{1}" display="inline"><mml:msub><mml:mi>F</mml:mi><mml:mn>1</mml:mn></mml:msub></mml:math></inline-formula> to obtain <inline-formula><mml:math alttext="F_{1}^{\prime}" display="inline"><mml:msubsup><mml:mi>F</mml:mi><mml:mn>1</mml:mn><mml:mo>′</mml:mo></mml:msubsup></mml:math></inline-formula>. Secondly, we calculate the Spatial attention weights <inline-formula><mml:math alttext="M_{S}" display="inline"><mml:msub><mml:mi>M</mml:mi><mml:mi>S</mml:mi></mml:msub></mml:math></inline-formula> of the feature map <inline-formula><mml:math alttext="F_{2}" display="inline"><mml:msub><mml:mi>F</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math></inline-formula> and multiply it with the feature map <inline-formula><mml:math alttext="F_{2}" display="inline"><mml:msub><mml:mi>F</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math></inline-formula> to obtain <inline-formula><mml:math alttext="F_{2}^{\prime}" display="inline"><mml:msubsup><mml:mi>F</mml:mi><mml:mn>2</mml:mn><mml:mo>′</mml:mo></mml:msubsup></mml:math></inline-formula>. Thirdly, we concatenate the feature map <inline-formula><mml:math alttext="F_{1}^{\prime}" display="inline"><mml:msubsup><mml:mi>F</mml:mi><mml:mn>1</mml:mn><mml:mo>′</mml:mo></mml:msubsup></mml:math></inline-formula> and <inline-formula><mml:math alttext="F_{2}^{\prime}" display="inline"><mml:msubsup><mml:mi>F</mml:mi><mml:mn>2</mml:mn><mml:mo>′</mml:mo></mml:msubsup></mml:math></inline-formula> to obtain the final feature map <inline-formula><mml:math alttext="F_{final}" display="inline"><mml:msub><mml:mi>F</mml:mi><mml:mrow><mml:mi>f</mml:mi><mml:mo>⁢</mml:mo><mml:mi>i</mml:mi><mml:mo>⁢</mml:mo><mml:mi>n</mml:mi><mml:mo>⁢</mml:mo><mml:mi>a</mml:mi><mml:mo>⁢</mml:mo><mml:mi>l</mml:mi></mml:mrow></mml:msub></mml:math></inline-formula>. The detailed process is as follows:</p>
        <p>
          <disp-formula id="S3.E1">
            <mml:math alttext="M_{C}=\sigma\left(MLP\left(\operatorname{AvgPool}\left(F_{1}\right)\right)+MLP%&#10;\left(\operatorname{MaxPool}\left(F_{1}\right)\right)\right)" display="inline">
              <mml:mrow>
                <mml:msub>
                  <mml:mi>M</mml:mi>
                  <mml:mi>C</mml:mi>
                </mml:msub>
                <mml:mo>=</mml:mo>
                <mml:mrow>
                  <mml:mi>σ</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mrow>
                    <mml:mo>(</mml:mo>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mi>M</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>L</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>P</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mrow>
                          <mml:mo>(</mml:mo>
                          <mml:mrow>
                            <mml:mi>AvgPool</mml:mi>
                            <mml:mo>⁡</mml:mo>
                            <mml:mrow>
                              <mml:mo>(</mml:mo>
                              <mml:msub>
                                <mml:mi>F</mml:mi>
                                <mml:mn>1</mml:mn>
                              </mml:msub>
                              <mml:mo>)</mml:mo>
                            </mml:mrow>
                          </mml:mrow>
                          <mml:mo>)</mml:mo>
                        </mml:mrow>
                      </mml:mrow>
                      <mml:mo>+</mml:mo>
                      <mml:mrow>
                        <mml:mi>M</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>L</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>P</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mrow>
                          <mml:mo>(</mml:mo>
                          <mml:mrow>
                            <mml:mi>MaxPool</mml:mi>
                            <mml:mo>⁡</mml:mo>
                            <mml:mrow>
                              <mml:mo>(</mml:mo>
                              <mml:msub>
                                <mml:mi>F</mml:mi>
                                <mml:mn>1</mml:mn>
                              </mml:msub>
                              <mml:mo>)</mml:mo>
                            </mml:mrow>
                          </mml:mrow>
                          <mml:mo>)</mml:mo>
                        </mml:mrow>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mo>)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p>
          <disp-formula id="S3.E2">
            <mml:math alttext="F_{1}^{\prime}=M_{C}\otimes F_{1}" display="block">
              <mml:mrow>
                <mml:msubsup>
                  <mml:mi>F</mml:mi>
                  <mml:mn>1</mml:mn>
                  <mml:mo>′</mml:mo>
                </mml:msubsup>
                <mml:mo>=</mml:mo>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>M</mml:mi>
                    <mml:mi>C</mml:mi>
                  </mml:msub>
                  <mml:mo lspace="0.222em" rspace="0.222em">⊗</mml:mo>
                  <mml:msub>
                    <mml:mi>F</mml:mi>
                    <mml:mn>1</mml:mn>
                  </mml:msub>
                </mml:mrow>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p>
          <disp-formula id="S3.E3">
            <mml:math alttext="M_{S}=\sigma(Conv(AvgPool(F_{2});MaxPool(F_{2})))" display="block">
              <mml:mrow>
                <mml:msub>
                  <mml:mi>M</mml:mi>
                  <mml:mi>S</mml:mi>
                </mml:msub>
                <mml:mo>=</mml:mo>
                <mml:mrow>
                  <mml:mi>σ</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mrow>
                    <mml:mo stretchy="false">(</mml:mo>
                    <mml:mrow>
                      <mml:mi>C</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>o</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>n</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>v</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mrow>
                        <mml:mo stretchy="false">(</mml:mo>
                        <mml:mrow>
                          <mml:mi>A</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>v</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>g</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>P</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>l</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mrow>
                            <mml:mo stretchy="false">(</mml:mo>
                            <mml:msub>
                              <mml:mi>F</mml:mi>
                              <mml:mn>2</mml:mn>
                            </mml:msub>
                            <mml:mo stretchy="false">)</mml:mo>
                          </mml:mrow>
                        </mml:mrow>
                        <mml:mo>;</mml:mo>
                        <mml:mrow>
                          <mml:mi>M</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>a</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>x</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>P</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>o</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>l</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mrow>
                            <mml:mo stretchy="false">(</mml:mo>
                            <mml:msub>
                              <mml:mi>F</mml:mi>
                              <mml:mn>2</mml:mn>
                            </mml:msub>
                            <mml:mo stretchy="false">)</mml:mo>
                          </mml:mrow>
                        </mml:mrow>
                        <mml:mo stretchy="false">)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mo stretchy="false">)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p>
          <disp-formula id="S3.E4">
            <mml:math alttext="F_{2}^{\prime}=M_{S}\otimes F_{2}" display="block">
              <mml:mrow>
                <mml:msubsup>
                  <mml:mi>F</mml:mi>
                  <mml:mn>2</mml:mn>
                  <mml:mo>′</mml:mo>
                </mml:msubsup>
                <mml:mo>=</mml:mo>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>M</mml:mi>
                    <mml:mi>S</mml:mi>
                  </mml:msub>
                  <mml:mo lspace="0.222em" rspace="0.222em">⊗</mml:mo>
                  <mml:msub>
                    <mml:mi>F</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:msub>
                </mml:mrow>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p>
          <disp-formula id="S3.E5">
            <mml:math alttext="F_{final}=Concat(F_{1}^{\prime};F_{2}^{\prime})" display="block">
              <mml:mrow>
                <mml:msub>
                  <mml:mi>F</mml:mi>
                  <mml:mrow>
                    <mml:mi>f</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>i</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>n</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>a</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>l</mml:mi>
                  </mml:mrow>
                </mml:msub>
                <mml:mo>=</mml:mo>
                <mml:mrow>
                  <mml:mi>C</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>o</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>n</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>c</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>a</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>t</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mrow>
                    <mml:mo stretchy="false">(</mml:mo>
                    <mml:msubsup>
                      <mml:mi>F</mml:mi>
                      <mml:mn>1</mml:mn>
                      <mml:mo>′</mml:mo>
                    </mml:msubsup>
                    <mml:mo>;</mml:mo>
                    <mml:msubsup>
                      <mml:mi>F</mml:mi>
                      <mml:mn>2</mml:mn>
                      <mml:mo>′</mml:mo>
                    </mml:msubsup>
                    <mml:mo stretchy="false">)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p id="S3.SS1.p5">where the feature map <inline-formula><mml:math alttext="F_{1}" display="inline"><mml:msub><mml:mi>F</mml:mi><mml:mn>1</mml:mn></mml:msub></mml:math></inline-formula> is derived from deeper layers and contains more semantic information, such as the overall shape and categories of sea ice; the feature map <inline-formula><mml:math alttext="F_{2}" display="inline"><mml:msub><mml:mi>F</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math></inline-formula> is derived from shallower layers and includes more detailed information, such as edges, textures, colors, and other low-level features of sea ice. In this way, we optimize the fusion process of different feature maps in YOLOv8, enabling the model to balance attention between the semantic information and detailed information of different feature maps, thereby improving the detection accuracy of the model for sea ice of varying scales.</p>
      </sec>
      <sec id="S3.SS2">
        <label>3.2</label>
        <title>Selection of Boundary Regression Loss Function Based on Sea Ice Size Characteristics</title>
        <p id="S3.SS2.p1">The C-IoU loss [<xref rid="ref050" ref-type="bibr">50</xref>] employed in YOLOv8 considers the geometric relationship between the ground truth box and the predicted box, utilizing both their relative positions and shapes to compute the loss. However, in contrast to general objects, sea ice presents a more diverse aspect ratio and possesses an irregular shape that lacks any fixed pattern. In this study, if C-IoU [<xref rid="ref050" ref-type="bibr">50</xref>] is utilized as the loss function for bounding box regression, two critical questions arise.</p>
        <p>
          <list list-type="order" id="S3.I1">
            <list-item id="S3.I1.i1">
              <p id="S3.I1.i1.p1">As shown in Figure <xref ref-type="fig" rid="F5">5</xref>, targets \scriptsize1⃝ and \scriptsize2⃝ represent the same sea ice. When two bounding boxes exhibit the same absolute deviation from the ground truth box, the bounding box that regresses from the direction of the shorter side of the rectangle tends to demonstrate a lower Intersection over Union (IoU) value. Our research indicates that this variation in IoU is more pronounced when regression occurs from the direction of the shorter side during bounding box adjustment. Consequently, it is crucial for models to effectively balance the regression impact of bounding boxes originating from different directions.</p>
              <p>
                <fig id="F5">
                  <label>Figure 5.</label>
                  <caption>
                    <p>Schematic diagram of the Question 1.</p>
                  </caption>
                  <graphic xlink:href="fig5Q1.pdf"/>
                </fig>
              </p>
            </list-item>
            <list-item id="S3.I1.i2">
              <p id="S3.I1.i2.p1">As illustrated in Figure <xref ref-type="fig" rid="F6">6</xref>, the center points of the prediction boxes \scriptsize3⃝ and \scriptsize4⃝ have shifted closer to the position of the ground truth box. Furthermore, both prediction boxes maintain an equal distance from the ground truth box along both the long and short sides. However, it is noteworthy that target \scriptsize4⃝, which regressed from the short side, corresponds to a lower IoU value, indicating reduced overlap with the ground truth.</p>
              <p>
                <fig id="F6">
                  <label>Figure 6.</label>
                  <caption>
                    <p>Schematic diagram of the Question 2.</p>
                  </caption>
                  <graphic xlink:href="fig6Q2.pdf"/>
                </fig>
              </p>
            </list-item>
          </list>
        </p>
        <p id="S3.SS2.p3">During the bounding box regression process, the variation in IoU is particularly significant when the regression occurs along the short side of the ground truth box. Therefore, it is essential to ensure a balanced regression effect for bounding boxes across various directions throughout this process.</p>
        <p id="S3.SS2.p4">In the end, we select Shape-IoU [<xref rid="ref041" ref-type="bibr">41</xref>] as our loss function for bounding box regression, as it effectively addresses the two types of issues mentioned above. The formula for Shape-IoU [<xref rid="ref041" ref-type="bibr">41</xref>] is presented below:</p>
        <p>
          <disp-formula id="S3.E6">
            <mml:math alttext="\left\{\begin{matrix}ww=\frac{2\ast(w_{gt})^{scale}}{(w_{gt})^{scale}+(h_{gt})%&#10;^{scale}}\\&#10;hh=\frac{2\ast(h_{gt})^{scale}}{(w_{gt})^{scale}+(h_{gt})^{scale}}\end{matrix}\right." class="ltx_math_unparsed" display="block">
              <mml:mrow>
                <mml:mo>{</mml:mo>
                <mml:mtable displaystyle="true" rowspacing="0pt">
                  <mml:mtr>
                    <mml:mtd>
                      <mml:mrow>
                        <mml:mrow>
                          <mml:mi>w</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>w</mml:mi>
                        </mml:mrow>
                        <mml:mo>=</mml:mo>
                        <mml:mstyle displaystyle="false">
                          <mml:mfrac>
                            <mml:mrow>
                              <mml:mn>2</mml:mn>
                              <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                              <mml:msup>
                                <mml:mrow>
                                  <mml:mo stretchy="false">(</mml:mo>
                                  <mml:msub>
                                    <mml:mi>w</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                  <mml:mo stretchy="false">)</mml:mo>
                                </mml:mrow>
                                <mml:mrow>
                                  <mml:mi>s</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>a</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>e</mml:mi>
                                </mml:mrow>
                              </mml:msup>
                            </mml:mrow>
                            <mml:mrow>
                              <mml:msup>
                                <mml:mrow>
                                  <mml:mo stretchy="false">(</mml:mo>
                                  <mml:msub>
                                    <mml:mi>w</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                  <mml:mo stretchy="false">)</mml:mo>
                                </mml:mrow>
                                <mml:mrow>
                                  <mml:mi>s</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>a</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>e</mml:mi>
                                </mml:mrow>
                              </mml:msup>
                              <mml:mo>+</mml:mo>
                              <mml:msup>
                                <mml:mrow>
                                  <mml:mo stretchy="false">(</mml:mo>
                                  <mml:msub>
                                    <mml:mi>h</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                  <mml:mo stretchy="false">)</mml:mo>
                                </mml:mrow>
                                <mml:mrow>
                                  <mml:mi>s</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>a</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>e</mml:mi>
                                </mml:mrow>
                              </mml:msup>
                            </mml:mrow>
                          </mml:mfrac>
                        </mml:mstyle>
                      </mml:mrow>
                    </mml:mtd>
                  </mml:mtr>
                  <mml:mtr>
                    <mml:mtd>
                      <mml:mrow>
                        <mml:mrow>
                          <mml:mi>h</mml:mi>
                          <mml:mo>⁢</mml:mo>
                          <mml:mi>h</mml:mi>
                        </mml:mrow>
                        <mml:mo>=</mml:mo>
                        <mml:mstyle displaystyle="false">
                          <mml:mfrac>
                            <mml:mrow>
                              <mml:mn>2</mml:mn>
                              <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                              <mml:msup>
                                <mml:mrow>
                                  <mml:mo stretchy="false">(</mml:mo>
                                  <mml:msub>
                                    <mml:mi>h</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                  <mml:mo stretchy="false">)</mml:mo>
                                </mml:mrow>
                                <mml:mrow>
                                  <mml:mi>s</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>a</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>e</mml:mi>
                                </mml:mrow>
                              </mml:msup>
                            </mml:mrow>
                            <mml:mrow>
                              <mml:msup>
                                <mml:mrow>
                                  <mml:mo stretchy="false">(</mml:mo>
                                  <mml:msub>
                                    <mml:mi>w</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                  <mml:mo stretchy="false">)</mml:mo>
                                </mml:mrow>
                                <mml:mrow>
                                  <mml:mi>s</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>a</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>e</mml:mi>
                                </mml:mrow>
                              </mml:msup>
                              <mml:mo>+</mml:mo>
                              <mml:msup>
                                <mml:mrow>
                                  <mml:mo stretchy="false">(</mml:mo>
                                  <mml:msub>
                                    <mml:mi>h</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                  <mml:mo stretchy="false">)</mml:mo>
                                </mml:mrow>
                                <mml:mrow>
                                  <mml:mi>s</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>a</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>e</mml:mi>
                                </mml:mrow>
                              </mml:msup>
                            </mml:mrow>
                          </mml:mfrac>
                        </mml:mstyle>
                      </mml:mrow>
                    </mml:mtd>
                  </mml:mtr>
                </mml:mtable>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p>
          <disp-formula id="S3.E7">
            <mml:math alttext="distance_{shape}=hh\ast\frac{(x-x_{gt})^{2}}{W^{2}+H^{2}}+ww\ast\frac{(y-y_{gt%&#10;})^{2}}{W^{2}+H^{2}}" display="block">
              <mml:mrow>
                <mml:mrow>
                  <mml:mi>d</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>i</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>s</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>t</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>a</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>n</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:mi>c</mml:mi>
                  <mml:mo>⁢</mml:mo>
                  <mml:msub>
                    <mml:mi>e</mml:mi>
                    <mml:mrow>
                      <mml:mi>s</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>h</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>a</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>p</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>e</mml:mi>
                    </mml:mrow>
                  </mml:msub>
                </mml:mrow>
                <mml:mo>=</mml:mo>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mrow>
                      <mml:mi>h</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>h</mml:mi>
                    </mml:mrow>
                    <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                    <mml:mfrac>
                      <mml:msup>
                        <mml:mrow>
                          <mml:mo stretchy="false">(</mml:mo>
                          <mml:mrow>
                            <mml:mi>x</mml:mi>
                            <mml:mo>−</mml:mo>
                            <mml:msub>
                              <mml:mi>x</mml:mi>
                              <mml:mrow>
                                <mml:mi>g</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>t</mml:mi>
                              </mml:mrow>
                            </mml:msub>
                          </mml:mrow>
                          <mml:mo stretchy="false">)</mml:mo>
                        </mml:mrow>
                        <mml:mn>2</mml:mn>
                      </mml:msup>
                      <mml:mrow>
                        <mml:msup>
                          <mml:mi>W</mml:mi>
                          <mml:mn>2</mml:mn>
                        </mml:msup>
                        <mml:mo>+</mml:mo>
                        <mml:msup>
                          <mml:mi>H</mml:mi>
                          <mml:mn>2</mml:mn>
                        </mml:msup>
                      </mml:mrow>
                    </mml:mfrac>
                  </mml:mrow>
                  <mml:mo>+</mml:mo>
                  <mml:mrow>
                    <mml:mrow>
                      <mml:mi>w</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>w</mml:mi>
                    </mml:mrow>
                    <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                    <mml:mfrac>
                      <mml:msup>
                        <mml:mrow>
                          <mml:mo stretchy="false">(</mml:mo>
                          <mml:mrow>
                            <mml:mi>y</mml:mi>
                            <mml:mo>−</mml:mo>
                            <mml:msub>
                              <mml:mi>y</mml:mi>
                              <mml:mrow>
                                <mml:mi>g</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>t</mml:mi>
                              </mml:mrow>
                            </mml:msub>
                          </mml:mrow>
                          <mml:mo stretchy="false">)</mml:mo>
                        </mml:mrow>
                        <mml:mn>2</mml:mn>
                      </mml:msup>
                      <mml:mrow>
                        <mml:msup>
                          <mml:mi>W</mml:mi>
                          <mml:mn>2</mml:mn>
                        </mml:msup>
                        <mml:mo>+</mml:mo>
                        <mml:msup>
                          <mml:mi>H</mml:mi>
                          <mml:mn>2</mml:mn>
                        </mml:msup>
                      </mml:mrow>
                    </mml:mfrac>
                  </mml:mrow>
                </mml:mrow>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p>
          <disp-formula id="S3.E8">
            <mml:math alttext="\left\{\begin{matrix}\omega_{w}=hh\ast\frac{\left|w-w_{gt}\right|}{max(w,w_{gt%&#10;})}\\&#10;\omega_{h}=ww\ast\frac{\left|h-h_{gt}\right|}{max(h,h_{gt})}\end{matrix}\right." class="ltx_math_unparsed" display="block">
              <mml:mrow>
                <mml:mo>{</mml:mo>
                <mml:mtable displaystyle="true" rowspacing="0pt">
                  <mml:mtr>
                    <mml:mtd>
                      <mml:mrow>
                        <mml:msub>
                          <mml:mi>ω</mml:mi>
                          <mml:mi>w</mml:mi>
                        </mml:msub>
                        <mml:mo>=</mml:mo>
                        <mml:mrow>
                          <mml:mrow>
                            <mml:mi>h</mml:mi>
                            <mml:mo>⁢</mml:mo>
                            <mml:mi>h</mml:mi>
                          </mml:mrow>
                          <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                          <mml:mstyle displaystyle="false">
                            <mml:mfrac>
                              <mml:mrow>
                                <mml:mo>|</mml:mo>
                                <mml:mrow>
                                  <mml:mi>w</mml:mi>
                                  <mml:mo>−</mml:mo>
                                  <mml:msub>
                                    <mml:mi>w</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                </mml:mrow>
                                <mml:mo>|</mml:mo>
                              </mml:mrow>
                              <mml:mrow>
                                <mml:mi>m</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>a</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>x</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mrow>
                                  <mml:mo stretchy="false">(</mml:mo>
                                  <mml:mi>w</mml:mi>
                                  <mml:mo>,</mml:mo>
                                  <mml:msub>
                                    <mml:mi>w</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                  <mml:mo stretchy="false">)</mml:mo>
                                </mml:mrow>
                              </mml:mrow>
                            </mml:mfrac>
                          </mml:mstyle>
                        </mml:mrow>
                      </mml:mrow>
                    </mml:mtd>
                  </mml:mtr>
                  <mml:mtr>
                    <mml:mtd>
                      <mml:mrow>
                        <mml:msub>
                          <mml:mi>ω</mml:mi>
                          <mml:mi>h</mml:mi>
                        </mml:msub>
                        <mml:mo>=</mml:mo>
                        <mml:mrow>
                          <mml:mrow>
                            <mml:mi>w</mml:mi>
                            <mml:mo>⁢</mml:mo>
                            <mml:mi>w</mml:mi>
                          </mml:mrow>
                          <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                          <mml:mstyle displaystyle="false">
                            <mml:mfrac>
                              <mml:mrow>
                                <mml:mo>|</mml:mo>
                                <mml:mrow>
                                  <mml:mi>h</mml:mi>
                                  <mml:mo>−</mml:mo>
                                  <mml:msub>
                                    <mml:mi>h</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                </mml:mrow>
                                <mml:mo>|</mml:mo>
                              </mml:mrow>
                              <mml:mrow>
                                <mml:mi>m</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>a</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>x</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mrow>
                                  <mml:mo stretchy="false">(</mml:mo>
                                  <mml:mi>h</mml:mi>
                                  <mml:mo>,</mml:mo>
                                  <mml:msub>
                                    <mml:mi>h</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>t</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                  <mml:mo stretchy="false">)</mml:mo>
                                </mml:mrow>
                              </mml:mrow>
                            </mml:mfrac>
                          </mml:mstyle>
                        </mml:mrow>
                      </mml:mrow>
                    </mml:mtd>
                  </mml:mtr>
                </mml:mtable>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p>
          <disp-formula id="S3.E9">
            <mml:math alttext="\Omega_{shape}=\sum_{t=w,h}(1-e^{-\omega_{t}})^{\theta},\theta=4" display="block">
              <mml:mrow>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi mathvariant="normal">Ω</mml:mi>
                    <mml:mrow>
                      <mml:mi>s</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>h</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>a</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>p</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>e</mml:mi>
                    </mml:mrow>
                  </mml:msub>
                  <mml:mo rspace="0.111em">=</mml:mo>
                  <mml:mrow>
                    <mml:munder>
                      <mml:mo movablelimits="false" rspace="0em">∑</mml:mo>
                      <mml:mrow>
                        <mml:mi>t</mml:mi>
                        <mml:mo>=</mml:mo>
                        <mml:mrow>
                          <mml:mi>w</mml:mi>
                          <mml:mo>,</mml:mo>
                          <mml:mi>h</mml:mi>
                        </mml:mrow>
                      </mml:mrow>
                    </mml:munder>
                    <mml:msup>
                      <mml:mrow>
                        <mml:mo stretchy="false">(</mml:mo>
                        <mml:mrow>
                          <mml:mn>1</mml:mn>
                          <mml:mo>−</mml:mo>
                          <mml:msup>
                            <mml:mi>e</mml:mi>
                            <mml:mrow>
                              <mml:mo>−</mml:mo>
                              <mml:msub>
                                <mml:mi>ω</mml:mi>
                                <mml:mi>t</mml:mi>
                              </mml:msub>
                            </mml:mrow>
                          </mml:msup>
                        </mml:mrow>
                        <mml:mo stretchy="false">)</mml:mo>
                      </mml:mrow>
                      <mml:mi>θ</mml:mi>
                    </mml:msup>
                  </mml:mrow>
                </mml:mrow>
                <mml:mo>,</mml:mo>
                <mml:mrow>
                  <mml:mi>θ</mml:mi>
                  <mml:mo>=</mml:mo>
                  <mml:mn>4</mml:mn>
                </mml:mrow>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p id="S3.SS2.p5">where <inline-formula><mml:math alttext="scale" display="inline"><mml:mrow><mml:mi>s</mml:mi><mml:mo>⁢</mml:mo><mml:mi>c</mml:mi><mml:mo>⁢</mml:mo><mml:mi>a</mml:mi><mml:mo>⁢</mml:mo><mml:mi>l</mml:mi><mml:mo>⁢</mml:mo><mml:mi>e</mml:mi></mml:mrow></mml:math></inline-formula> represents the scale factor, which can be adjusted based on the dimensions of the target. Taking the ground truth box in Figure <xref ref-type="fig" rid="F5">5</xref> as an example(where <inline-formula><mml:math alttext="w_{gt}&gt;h_{gt}" display="inline"><mml:mrow><mml:msub><mml:mi>w</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mo>⁢</mml:mo><mml:mi>t</mml:mi></mml:mrow></mml:msub><mml:mo>&gt;</mml:mo><mml:msub><mml:mi>h</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mo>⁢</mml:mo><mml:mi>t</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula>), when <inline-formula><mml:math alttext="scale=0" display="inline"><mml:mrow><mml:mrow><mml:mi>s</mml:mi><mml:mo>⁢</mml:mo><mml:mi>c</mml:mi><mml:mo>⁢</mml:mo><mml:mi>a</mml:mi><mml:mo>⁢</mml:mo><mml:mi>l</mml:mi><mml:mo>⁢</mml:mo><mml:mi>e</mml:mi></mml:mrow><mml:mo>=</mml:mo><mml:mn>0</mml:mn></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math alttext="ww=hh=1" display="inline"><mml:mrow><mml:mrow><mml:mi>w</mml:mi><mml:mo>⁢</mml:mo><mml:mi>w</mml:mi></mml:mrow><mml:mo>=</mml:mo><mml:mrow><mml:mi>h</mml:mi><mml:mo>⁢</mml:mo><mml:mi>h</mml:mi></mml:mrow><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:math></inline-formula>, the bounding box regression lacks directional prioritization. by increasing the value of scale, the regression effectiveness can be enhanced. In this study, we set <inline-formula><mml:math alttext="scale=1" display="inline"><mml:mrow><mml:mrow><mml:mi>s</mml:mi><mml:mo>⁢</mml:mo><mml:mi>c</mml:mi><mml:mo>⁢</mml:mo><mml:mi>a</mml:mi><mml:mo>⁢</mml:mo><mml:mi>l</mml:mi><mml:mo>⁢</mml:mo><mml:mi>e</mml:mi></mml:mrow><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:math></inline-formula>, resulting in <inline-formula><mml:math alttext="ww&gt;hh" display="inline"><mml:mrow><mml:mrow><mml:mi>w</mml:mi><mml:mo>⁢</mml:mo><mml:mi>w</mml:mi></mml:mrow><mml:mo>&gt;</mml:mo><mml:mrow><mml:mi>h</mml:mi><mml:mo>⁢</mml:mo><mml:mi>h</mml:mi></mml:mrow></mml:mrow></mml:math></inline-formula>, which indicates that higher regression weight is assigned to the vertical dimension (height adjustment).</p>
        <p id="S3.SS2.p6">Equation <xref rid="S3.E10">10</xref> calculates the loss value for bounding box regression.</p>
        <p>
          <disp-formula id="S3.E10">
            <mml:math alttext="L_{Shape-IoU}=1-IoU+distance_{shape}+0.5\ast\Omega_{shape}" display="block">
              <mml:mrow>
                <mml:msub>
                  <mml:mi>L</mml:mi>
                  <mml:mrow>
                    <mml:mrow>
                      <mml:mi>S</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>h</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>a</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>p</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>e</mml:mi>
                    </mml:mrow>
                    <mml:mo>−</mml:mo>
                    <mml:mrow>
                      <mml:mi>I</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>o</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>U</mml:mi>
                    </mml:mrow>
                  </mml:mrow>
                </mml:msub>
                <mml:mo>=</mml:mo>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mn>1</mml:mn>
                    <mml:mo>−</mml:mo>
                    <mml:mrow>
                      <mml:mi>I</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>o</mml:mi>
                      <mml:mo>⁢</mml:mo>
                      <mml:mi>U</mml:mi>
                    </mml:mrow>
                  </mml:mrow>
                  <mml:mo>+</mml:mo>
                  <mml:mrow>
                    <mml:mi>d</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>i</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>s</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>t</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>a</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>n</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:mi>c</mml:mi>
                    <mml:mo>⁢</mml:mo>
                    <mml:msub>
                      <mml:mi>e</mml:mi>
                      <mml:mrow>
                        <mml:mi>s</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>h</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>a</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>p</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>e</mml:mi>
                      </mml:mrow>
                    </mml:msub>
                  </mml:mrow>
                  <mml:mo>+</mml:mo>
                  <mml:mrow>
                    <mml:mn>0.5</mml:mn>
                    <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                    <mml:msub>
                      <mml:mi mathvariant="normal">Ω</mml:mi>
                      <mml:mrow>
                        <mml:mi>s</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>h</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>a</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>p</mml:mi>
                        <mml:mo>⁢</mml:mo>
                        <mml:mi>e</mml:mi>
                      </mml:mrow>
                    </mml:msub>
                  </mml:mrow>
                </mml:mrow>
              </mml:mrow>
            </mml:math>
          </disp-formula>
        </p>
        <p id="S3.SS2.p7">It is noteworthy that Shape-IoU dynamically emphasizes the gradient update path of bounding box parameters (e.g., center offsets and aspect ratios) during model convergence. Unlike C-IoU, which indirectly guides optimization through geometric penalties (center distance and aspect ratio matching), Shape-IoU explicitly introduces a directional weighting coefficient, thereby clarifying the prioritization of regression targets with higher shape discrepancies. This property of Shape-IoU shortens the convergence time of the model.</p>
      </sec>
      <sec id="S3.SS3">
        <label>3.3</label>
        <title>An Evidence Fusion Module for the Correction of Sea Ice Categories</title>
        <p>
          <fig id="F7">
            <label>Figure 7.</label>
            <caption>
              <p>Typical cases of misclassification.</p>
            </caption>
            <graphic xlink:href="fig7Q.pdf"/>
          </fig>
        </p>
        <p>
          <fig id="F8">
            <label>Figure 8.</label>
            <caption>
              <p>Schematic diagram of evidence fusion module.</p>
            </caption>
            <graphic xlink:href="fig8.pdf"/>
          </fig>
        </p>
        <p id="S3.SS3.p1">After extensive experimentation, we discovered that YOLOv8 is capable of accurately predicting the bounding boxes of sea ice; however, it occasionally misclassifies the categories of sea ice. More specifically, YOLOv8 occasionally misclassifies medium-scale sea ice as large-scale sea ice and conversely misclassifies large-scale sea ice as medium-scale sea ice, As illustrated in Figure <xref ref-type="fig" rid="F7">7</xref>.</p>
        <p id="S3.SS3.p2">In Figure <xref ref-type="fig" rid="F7">7</xref> (a), the sea ice located in the upper left corner is classified as medium-scale, as the longest side of its circumscribed rectangle measures less than 128 pixels. However, it was incorrectly identified by YOLOv8 as large sea ice. Meanwhile, in Figure <xref ref-type="fig" rid="F7">7</xref> (b), the sea ice situated at the center is categorized as large-scale, given that the longest side of its circumscribed rectangle exceeds 128 pixels. However, it was inaccurately classified by YOLOv8 as medium-scale sea ice.</p>
        <p id="S3.SS3.p3">With these issues in consideration, we conducted a more thorough examination of YOLO. The YOLO algorithm is designed to extract features from images and classify targets based on these extracted characteristics. The features encompass various types of information, including texture, color, shape, and more. More specifically, YOLO relies more heavily on the aforementioned features for target classification than on the scale information of the targets.</p>
        <p id="S3.SS3.p4">However, in our task of classifying sea ice, the scale information of the target cannot be overlooked. We aim to improve YOLOv8 so that the scale information of the target can serve as a more significant feature for predicting categories of sea ice.</p>
        <p id="S3.SS3.p5">In the inference process of YOLOv8, the role of non-maximum suppression (NMS) is to eliminate redundant prediction boxes and produce the final output. Based on this, we propose the Evidence Fusion module to address the aforementioned issues, the details of the Evidence Fusion module are illustrated in Figure <xref ref-type="fig" rid="F8">8</xref>.</p>
        <p id="S3.SS3.p6">First, we convert the prediction box information and prediction category information provided by YOLOv8 into evidences. Utilizing an enhanced DSmT fusion inference algorithm [<xref rid="ref042" ref-type="bibr">42</xref>], we subsequently integrate these two types of evidence to establish a new prediction category. The algorithmic model is primarily composed of the following two components.</p>
        <sec id="S3.SS3.SSS1">
          <label>3.3.1</label>
          <title>Convert the Information Predicted by YOLOv8 into Evidence Characterizing Uncertainty</title>
          <p>
            <list list-type="order" id="S3.I2">
              <list-item id="S3.I2.i1">
                <p id="S3.I2.i1.p1">The bounding box information predicted by YOLOv8: We begin by counting the instances of sea ice larger than 8 pixels in the satellite image dataset NWPU-RESISC45 [<xref rid="ref029" ref-type="bibr">29</xref>]. Subsequently, we categorize these sea ice instances into three distinct groups based on their scale, as elaborated in section <xref rid="S4.SS1">4.1</xref>. Finally, we generate histograms to illustrate the frequency distribution of the longest side of the circumscribed rectangles for each type of sea ice and fit distribution curves to these histograms, as depicted in Figure <xref ref-type="fig" rid="F9">9</xref>.</p>
                <p>
                  <fig id="F9">
                    <label>Figure 9.</label>
                    <caption>
                      <p>Distribution of the size of three types of sea ice in the NWPU-RESISC45 dataset: (a), (b), and (c) Frequency histograms and probability density curves showing the distribution of <italic>Small-scale sea ice</italic>, <italic>Medium-scale sea ice</italic> and <italic>Large-scale sea ice</italic>, respectively; (d) A schematic diagram is presented, showing the three curves.</p>
                    </caption>
                    <graphic xlink:href="fig9.pdf"/>
                  </fig>
                </p>
                <p id="S3.I2.i1.p2">Based on the distribution shown in Figure <xref ref-type="fig" rid="F9">9</xref>, we fit the curve, as shown in equation <xref rid="S3.E11">11</xref>, <xref rid="S3.E12">12</xref>, and <xref rid="S3.E13">13</xref>.</p>
                <p>
                  <disp-formula id="S3.E11">
                    <mml:math alttext="f_{small}(l)=\frac{1}{7.32\ast\sqrt{2\pi}}\ast e^{-\frac{(l-21.42)^{2}}{2\ast 7%&#10;.32^{2}}},l&gt;0" display="block">
                      <mml:mrow>
                        <mml:mrow>
                          <mml:mrow>
                            <mml:msub>
                              <mml:mi>f</mml:mi>
                              <mml:mrow>
                                <mml:mi>s</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>m</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>a</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>l</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>l</mml:mi>
                              </mml:mrow>
                            </mml:msub>
                            <mml:mo>⁢</mml:mo>
                            <mml:mrow>
                              <mml:mo stretchy="false">(</mml:mo>
                              <mml:mi>l</mml:mi>
                              <mml:mo stretchy="false">)</mml:mo>
                            </mml:mrow>
                          </mml:mrow>
                          <mml:mo>=</mml:mo>
                          <mml:mrow>
                            <mml:mfrac>
                              <mml:mn>1</mml:mn>
                              <mml:mrow>
                                <mml:mn>7.32</mml:mn>
                                <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                                <mml:msqrt>
                                  <mml:mrow>
                                    <mml:mn>2</mml:mn>
                                    <mml:mo>⁢</mml:mo>
                                    <mml:mi>π</mml:mi>
                                  </mml:mrow>
                                </mml:msqrt>
                              </mml:mrow>
                            </mml:mfrac>
                            <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                            <mml:msup>
                              <mml:mi>e</mml:mi>
                              <mml:mrow>
                                <mml:mo>−</mml:mo>
                                <mml:mfrac>
                                  <mml:msup>
                                    <mml:mrow>
                                      <mml:mo stretchy="false">(</mml:mo>
                                      <mml:mrow>
                                        <mml:mi>l</mml:mi>
                                        <mml:mo>−</mml:mo>
                                        <mml:mn>21.42</mml:mn>
                                      </mml:mrow>
                                      <mml:mo stretchy="false">)</mml:mo>
                                    </mml:mrow>
                                    <mml:mn>2</mml:mn>
                                  </mml:msup>
                                  <mml:mrow>
                                    <mml:mn>2</mml:mn>
                                    <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                                    <mml:msup>
                                      <mml:mn>7.32</mml:mn>
                                      <mml:mn>2</mml:mn>
                                    </mml:msup>
                                  </mml:mrow>
                                </mml:mfrac>
                              </mml:mrow>
                            </mml:msup>
                          </mml:mrow>
                        </mml:mrow>
                        <mml:mo>,</mml:mo>
                        <mml:mrow>
                          <mml:mi>l</mml:mi>
                          <mml:mo>&gt;</mml:mo>
                          <mml:mn>0</mml:mn>
                        </mml:mrow>
                      </mml:mrow>
                    </mml:math>
                  </disp-formula>
                </p>
                <p id="S3.I2.i1.p3">where 21.42 represents the mean <inline-formula><mml:math alttext="\mu" display="inline"><mml:mi>μ</mml:mi></mml:math></inline-formula> of the normal distribution, 7.32 represents the standard deviation <inline-formula><mml:math alttext="\sigma" display="inline"><mml:mi>σ</mml:mi></mml:math></inline-formula> of the normal distribution. Here, <inline-formula><mml:math alttext="l" display="inline"><mml:mi>l</mml:mi></mml:math></inline-formula> denotes the pixel value corresponding to the longest side of the circumscribed rectangle for sea ice, and <italic>f<sub>small</sub>(l)</italic> is the probability of occurrence of sea ice.</p>
                <p>
                  <disp-formula-group id="S3.E12">
                    <disp-formula id="S3.E12X">
                      <mml:math alttext="\displaystyle f_{medium}(l)={}~{}0.64\ast\frac{1}{2^{\frac{11.43}{2}}\ast%&#10;\Gamma\left(\frac{11.43}{2}\right)}" display="inline">
                        <mml:mrow>
                          <mml:mrow>
                            <mml:msub>
                              <mml:mi>f</mml:mi>
                              <mml:mrow>
                                <mml:mi>m</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>e</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>d</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>i</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>u</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>m</mml:mi>
                              </mml:mrow>
                            </mml:msub>
                            <mml:mo>⁢</mml:mo>
                            <mml:mrow>
                              <mml:mo stretchy="false">(</mml:mo>
                              <mml:mi>l</mml:mi>
                              <mml:mo stretchy="false">)</mml:mo>
                            </mml:mrow>
                          </mml:mrow>
                          <mml:mo rspace="0.608em">=</mml:mo>
                          <mml:mrow>
                            <mml:mn>0.64</mml:mn>
                            <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                            <mml:mstyle displaystyle="true">
                              <mml:mfrac>
                                <mml:mn>1</mml:mn>
                                <mml:mrow>
                                  <mml:mrow>
                                    <mml:msup>
                                      <mml:mn>2</mml:mn>
                                      <mml:mfrac>
                                        <mml:mn>11.43</mml:mn>
                                        <mml:mn>2</mml:mn>
                                      </mml:mfrac>
                                    </mml:msup>
                                    <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                                    <mml:mi mathvariant="normal">Γ</mml:mi>
                                  </mml:mrow>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mrow>
                                    <mml:mo>(</mml:mo>
                                    <mml:mfrac>
                                      <mml:mn>11.43</mml:mn>
                                      <mml:mn>2</mml:mn>
                                    </mml:mfrac>
                                    <mml:mo>)</mml:mo>
                                  </mml:mrow>
                                </mml:mrow>
                              </mml:mfrac>
                            </mml:mstyle>
                          </mml:mrow>
                        </mml:mrow>
                      </mml:math>
                    </disp-formula>
                    <disp-formula id="S3.E12Xa">
                      <mml:math alttext="\displaystyle\ast\left(\frac{l}{6.78}\right)^{\left(\frac{11.43}{2}-1\right)}%&#10;\ast e^{-\frac{l}{2\ast 6.78}},l&gt;0" class="ltx_math_unparsed" display="inline">
                        <mml:mrow>
                          <mml:mo rspace="0.222em">∗</mml:mo>
                          <mml:msup>
                            <mml:mrow>
                              <mml:mo>(</mml:mo>
                              <mml:mstyle displaystyle="true">
                                <mml:mfrac>
                                  <mml:mi>l</mml:mi>
                                  <mml:mn>6.78</mml:mn>
                                </mml:mfrac>
                              </mml:mstyle>
                              <mml:mo rspace="0.055em">)</mml:mo>
                            </mml:mrow>
                            <mml:mrow>
                              <mml:mo>(</mml:mo>
                              <mml:mrow>
                                <mml:mfrac>
                                  <mml:mn>11.43</mml:mn>
                                  <mml:mn>2</mml:mn>
                                </mml:mfrac>
                                <mml:mo>−</mml:mo>
                                <mml:mn>1</mml:mn>
                              </mml:mrow>
                              <mml:mo>)</mml:mo>
                            </mml:mrow>
                          </mml:msup>
                          <mml:mo rspace="0.222em">∗</mml:mo>
                          <mml:msup>
                            <mml:mi>e</mml:mi>
                            <mml:mrow>
                              <mml:mo>−</mml:mo>
                              <mml:mfrac>
                                <mml:mi>l</mml:mi>
                                <mml:mrow>
                                  <mml:mn>2</mml:mn>
                                  <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                                  <mml:mn>6.78</mml:mn>
                                </mml:mrow>
                              </mml:mfrac>
                            </mml:mrow>
                          </mml:msup>
                          <mml:mo>,</mml:mo>
                          <mml:mi>l</mml:mi>
                          <mml:mo>&gt;</mml:mo>
                          <mml:mn>0</mml:mn>
                        </mml:mrow>
                      </mml:math>
                    </disp-formula>
                  </disp-formula-group>
                </p>
                <p>
                  <fig-group>
                    <fig id="F10.a">
                      <caption>
                        <p>(a) Small-scale sea ice: the longest side of the circumscribed rectangle for this type of sea ice ranges between 8 pixels and 32 pixels.</p>
                        <graphic xlink:href="fig10(a)small.pdf"/>
                      </caption>
                    </fig>
                    <fig id="F10.b">
                      <caption>
                        <p>(b) Medium-scale sea ice: the longest side of the circumscribed rectangle for this type of sea ice ranges between 32 pixels and 128 pixels.</p>
                        <graphic xlink:href="fig10(b)medium.pdf"/>
                      </caption>
                    </fig>
                    <fig id="F10.c">
                      <caption>
                        <p>(c) Large-scale sea ice: the longest side of the circumscribed rectangle for this type of sea ice exceeds 128 pixels.</p>
                        <graphic xlink:href="fig10(c)big.pdf"/>
                      </caption>
                    </fig>
                    <fig id="F10">
                      <label>Figure 10.</label>
                      <caption>
                        <p>Three distinct categories of sea ice.</p>
                      </caption>
                    </fig>
                  </fig-group>
                </p>
                <p id="S3.I2.i1.p5">where 0.64 serves as the scaling parameter for the function, 6.78 is the scale parameter, 11.43 denotes the degrees of freedom for the chi-square distribution, and <inline-formula><mml:math alttext="\Gamma(\cdot)" display="inline"><mml:mrow><mml:mi mathvariant="normal">Γ</mml:mi><mml:mo>⁢</mml:mo><mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mo lspace="0em" rspace="0em">⋅</mml:mo><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> is the symbol for the gamma function. Here, <inline-formula><mml:math alttext="l" display="inline"><mml:mi>l</mml:mi></mml:math></inline-formula> denotes the pixel value corresponding to the longest side of the circumscribed rectangle for sea ice, and <italic>f<sub>medium</sub>(l)</italic> is the probability of occurrence of sea ice.</p>
                <p>
                  <disp-formula-group id="S3.E13">
                    <disp-formula id="S3.E13X">
                      <mml:math alttext="\displaystyle f_{large}(l)={}~{}1.06\ast\frac{1}{2^{\frac{69.06}{2}}\ast\Gamma%&#10;(\frac{69.06}{2})}" display="inline">
                        <mml:mrow>
                          <mml:mrow>
                            <mml:msub>
                              <mml:mi>f</mml:mi>
                              <mml:mrow>
                                <mml:mi>l</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>a</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>r</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>g</mml:mi>
                                <mml:mo>⁢</mml:mo>
                                <mml:mi>e</mml:mi>
                              </mml:mrow>
                            </mml:msub>
                            <mml:mo>⁢</mml:mo>
                            <mml:mrow>
                              <mml:mo stretchy="false">(</mml:mo>
                              <mml:mi>l</mml:mi>
                              <mml:mo stretchy="false">)</mml:mo>
                            </mml:mrow>
                          </mml:mrow>
                          <mml:mo rspace="0.608em">=</mml:mo>
                          <mml:mrow>
                            <mml:mn>1.06</mml:mn>
                            <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                            <mml:mstyle displaystyle="true">
                              <mml:mfrac>
                                <mml:mn>1</mml:mn>
                                <mml:mrow>
                                  <mml:mrow>
                                    <mml:msup>
                                      <mml:mn>2</mml:mn>
                                      <mml:mfrac>
                                        <mml:mn>69.06</mml:mn>
                                        <mml:mn>2</mml:mn>
                                      </mml:mfrac>
                                    </mml:msup>
                                    <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                                    <mml:mi mathvariant="normal">Γ</mml:mi>
                                  </mml:mrow>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mrow>
                                    <mml:mo stretchy="false">(</mml:mo>
                                    <mml:mfrac>
                                      <mml:mn>69.06</mml:mn>
                                      <mml:mn>2</mml:mn>
                                    </mml:mfrac>
                                    <mml:mo stretchy="false">)</mml:mo>
                                  </mml:mrow>
                                </mml:mrow>
                              </mml:mfrac>
                            </mml:mstyle>
                          </mml:mrow>
                        </mml:mrow>
                      </mml:math>
                    </disp-formula>
                    <disp-formula id="S3.E13Xa">
                      <mml:math alttext="\displaystyle\ast(\frac{l}{2.43})^{(\frac{69.06}{2}-1)}\ast e^{-\frac{l}{2\ast&#10;2%&#10;.43}},l&gt;0" class="ltx_math_unparsed" display="inline">
                        <mml:mrow>
                          <mml:mo rspace="0.222em">∗</mml:mo>
                          <mml:msup>
                            <mml:mrow>
                              <mml:mo stretchy="false">(</mml:mo>
                              <mml:mstyle displaystyle="true">
                                <mml:mfrac>
                                  <mml:mi>l</mml:mi>
                                  <mml:mn>2.43</mml:mn>
                                </mml:mfrac>
                              </mml:mstyle>
                              <mml:mo rspace="0.055em" stretchy="false">)</mml:mo>
                            </mml:mrow>
                            <mml:mrow>
                              <mml:mo stretchy="false">(</mml:mo>
                              <mml:mrow>
                                <mml:mfrac>
                                  <mml:mn>69.06</mml:mn>
                                  <mml:mn>2</mml:mn>
                                </mml:mfrac>
                                <mml:mo>−</mml:mo>
                                <mml:mn>1</mml:mn>
                              </mml:mrow>
                              <mml:mo stretchy="false">)</mml:mo>
                            </mml:mrow>
                          </mml:msup>
                          <mml:mo rspace="0.222em">∗</mml:mo>
                          <mml:msup>
                            <mml:mi>e</mml:mi>
                            <mml:mrow>
                              <mml:mo>−</mml:mo>
                              <mml:mfrac>
                                <mml:mi>l</mml:mi>
                                <mml:mrow>
                                  <mml:mn>2</mml:mn>
                                  <mml:mo lspace="0.222em" rspace="0.222em">∗</mml:mo>
                                  <mml:mn>2.43</mml:mn>
                                </mml:mrow>
                              </mml:mfrac>
                            </mml:mrow>
                          </mml:msup>
                          <mml:mo>,</mml:mo>
                          <mml:mi>l</mml:mi>
                          <mml:mo>&gt;</mml:mo>
                          <mml:mn>0</mml:mn>
                        </mml:mrow>
                      </mml:math>
                    </disp-formula>
                  </disp-formula-group>
                </p>
                <p id="S3.I2.i1.p6">where 1.06 serves as the scaling parameter for the function, 2.43 is the scale parameter, 69.06 denotes the degrees of freedom for the chi-square distribution, and <inline-formula><mml:math alttext="\Gamma(\cdot)" display="inline"><mml:mrow><mml:mi mathvariant="normal">Γ</mml:mi><mml:mo>⁢</mml:mo><mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mo lspace="0em" rspace="0em">⋅</mml:mo><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> is the symbol for the gamma function. Here, <inline-formula><mml:math alttext="l" display="inline"><mml:mi>l</mml:mi></mml:math></inline-formula> denotes the pixel value corresponding to the longest side of the circumscribed rectangle for sea ice, and <italic>f<sub>large</sub>(l)</italic> is the probability of occurrence of sea ice.</p>
                <p id="S3.I2.i1.p7">We normalize the distribution rules mentioned above, as shown in equations <xref rid="S3.E14">14</xref>, and finally transform the bounding box information predicted by YOLOv8 into evidence that describes the uncertainty.</p>
                <p>
                  <disp-formula id="S3.E14">
                    <mml:math alttext="\left\{\begin{matrix}a_{1}=\frac{f_{small}(l)}{f_{small}(l)+f_{middle}(l)+f_{%&#10;large}(l)}\\&#10;a_{2}=\frac{f_{medium}(l)}{f_{small}(l)+f_{medium}(l)+f_{large}(l)}\\&#10;a_{3}=\frac{f_{large}(l)}{f_{small}(l)+f_{medium}(l)+f_{large}(l)}\end{matrix}\right." class="ltx_math_unparsed" display="block">
                      <mml:mrow>
                        <mml:mo>{</mml:mo>
                        <mml:mtable displaystyle="true" rowspacing="0pt">
                          <mml:mtr>
                            <mml:mtd>
                              <mml:mrow>
                                <mml:msub>
                                  <mml:mi>a</mml:mi>
                                  <mml:mn>1</mml:mn>
                                </mml:msub>
                                <mml:mo>=</mml:mo>
                                <mml:mstyle displaystyle="false">
                                  <mml:mfrac>
                                    <mml:mrow>
                                      <mml:msub>
                                        <mml:mi>f</mml:mi>
                                        <mml:mrow>
                                          <mml:mi>s</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>m</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>a</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>l</mml:mi>
                                        </mml:mrow>
                                      </mml:msub>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mrow>
                                        <mml:mo stretchy="false">(</mml:mo>
                                        <mml:mi>l</mml:mi>
                                        <mml:mo stretchy="false">)</mml:mo>
                                      </mml:mrow>
                                    </mml:mrow>
                                    <mml:mrow>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>s</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>m</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>a</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>l</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>l</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                      <mml:mo>+</mml:mo>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>m</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>i</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>d</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>d</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>l</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>e</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                      <mml:mo>+</mml:mo>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>l</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>a</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>r</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>g</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>e</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                    </mml:mrow>
                                  </mml:mfrac>
                                </mml:mstyle>
                              </mml:mrow>
                            </mml:mtd>
                          </mml:mtr>
                          <mml:mtr>
                            <mml:mtd>
                              <mml:mrow>
                                <mml:msub>
                                  <mml:mi>a</mml:mi>
                                  <mml:mn>2</mml:mn>
                                </mml:msub>
                                <mml:mo>=</mml:mo>
                                <mml:mstyle displaystyle="false">
                                  <mml:mfrac>
                                    <mml:mrow>
                                      <mml:msub>
                                        <mml:mi>f</mml:mi>
                                        <mml:mrow>
                                          <mml:mi>m</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>e</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>d</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>i</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>u</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>m</mml:mi>
                                        </mml:mrow>
                                      </mml:msub>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mrow>
                                        <mml:mo stretchy="false">(</mml:mo>
                                        <mml:mi>l</mml:mi>
                                        <mml:mo stretchy="false">)</mml:mo>
                                      </mml:mrow>
                                    </mml:mrow>
                                    <mml:mrow>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>s</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>m</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>a</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>l</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>l</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                      <mml:mo>+</mml:mo>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>m</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>e</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>d</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>i</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>u</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>m</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                      <mml:mo>+</mml:mo>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>l</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>a</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>r</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>g</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>e</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                    </mml:mrow>
                                  </mml:mfrac>
                                </mml:mstyle>
                              </mml:mrow>
                            </mml:mtd>
                          </mml:mtr>
                          <mml:mtr>
                            <mml:mtd>
                              <mml:mrow>
                                <mml:msub>
                                  <mml:mi>a</mml:mi>
                                  <mml:mn>3</mml:mn>
                                </mml:msub>
                                <mml:mo>=</mml:mo>
                                <mml:mstyle displaystyle="false">
                                  <mml:mfrac>
                                    <mml:mrow>
                                      <mml:msub>
                                        <mml:mi>f</mml:mi>
                                        <mml:mrow>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>a</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>r</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>g</mml:mi>
                                          <mml:mo>⁢</mml:mo>
                                          <mml:mi>e</mml:mi>
                                        </mml:mrow>
                                      </mml:msub>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mrow>
                                        <mml:mo stretchy="false">(</mml:mo>
                                        <mml:mi>l</mml:mi>
                                        <mml:mo stretchy="false">)</mml:mo>
                                      </mml:mrow>
                                    </mml:mrow>
                                    <mml:mrow>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>s</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>m</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>a</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>l</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>l</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                      <mml:mo>+</mml:mo>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>m</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>e</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>d</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>i</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>u</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>m</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                      <mml:mo>+</mml:mo>
                                      <mml:mrow>
                                        <mml:msub>
                                          <mml:mi>f</mml:mi>
                                          <mml:mrow>
                                            <mml:mi>l</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>a</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>r</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>g</mml:mi>
                                            <mml:mo>⁢</mml:mo>
                                            <mml:mi>e</mml:mi>
                                          </mml:mrow>
                                        </mml:msub>
                                        <mml:mo>⁢</mml:mo>
                                        <mml:mrow>
                                          <mml:mo stretchy="false">(</mml:mo>
                                          <mml:mi>l</mml:mi>
                                          <mml:mo stretchy="false">)</mml:mo>
                                        </mml:mrow>
                                      </mml:mrow>
                                    </mml:mrow>
                                  </mml:mfrac>
                                </mml:mstyle>
                              </mml:mrow>
                            </mml:mtd>
                          </mml:mtr>
                        </mml:mtable>
                      </mml:mrow>
                    </mml:math>
                  </disp-formula>
                </p>
                <p id="S3.I2.i1.p8">where <italic>a<sub>1</sub></italic>, <italic>a<sub>2</sub></italic>, <italic>a<sub>3</sub></italic> represent the scale reliability for the three types of sea ice, respectively.</p>
              </list-item>
              <list-item id="S3.I2.i2">
                <p id="S3.I2.i2.p1">The category information predicted by YOLOv8: The prediction values for the three types of sea ice—<italic>cls<sub>small</sub></italic>, <italic>cls<sub>medium</sub></italic>, <italic>cls<sub>large</sub></italic>—are included in the prediction information provided by YOLOv8. The category with the highest prediction value indicates the model's predicted target. We convert this set of data into category evidence, as shown in equations <xref rid="S3.E15">15</xref>.</p>
                <p>
                  <disp-formula id="S3.E15">
                    <mml:math alttext="\left\{\begin{matrix}b_{1}=cls_{small}\\&#10;b_{2}=cls_{medium}\\&#10;b_{3}=cls_{large}\end{matrix}\right." class="ltx_math_unparsed" display="block">
                      <mml:mrow>
                        <mml:mo>{</mml:mo>
                        <mml:mtable displaystyle="true" rowspacing="0pt">
                          <mml:mtr>
                            <mml:mtd>
                              <mml:mrow>
                                <mml:msub>
                                  <mml:mi>b</mml:mi>
                                  <mml:mn>1</mml:mn>
                                </mml:msub>
                                <mml:mo>=</mml:mo>
                                <mml:mrow>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:msub>
                                    <mml:mi>s</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>s</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>m</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>a</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>l</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>l</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                </mml:mrow>
                              </mml:mrow>
                            </mml:mtd>
                          </mml:mtr>
                          <mml:mtr>
                            <mml:mtd>
                              <mml:mrow>
                                <mml:msub>
                                  <mml:mi>b</mml:mi>
                                  <mml:mn>2</mml:mn>
                                </mml:msub>
                                <mml:mo>=</mml:mo>
                                <mml:mrow>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:msub>
                                    <mml:mi>s</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>m</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>e</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>d</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>i</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>u</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>m</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                </mml:mrow>
                              </mml:mrow>
                            </mml:mtd>
                          </mml:mtr>
                          <mml:mtr>
                            <mml:mtd>
                              <mml:mrow>
                                <mml:msub>
                                  <mml:mi>b</mml:mi>
                                  <mml:mn>3</mml:mn>
                                </mml:msub>
                                <mml:mo>=</mml:mo>
                                <mml:mrow>
                                  <mml:mi>c</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:mi>l</mml:mi>
                                  <mml:mo>⁢</mml:mo>
                                  <mml:msub>
                                    <mml:mi>s</mml:mi>
                                    <mml:mrow>
                                      <mml:mi>l</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>a</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>r</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>g</mml:mi>
                                      <mml:mo>⁢</mml:mo>
                                      <mml:mi>e</mml:mi>
                                    </mml:mrow>
                                  </mml:msub>
                                </mml:mrow>
                              </mml:mrow>
                            </mml:mtd>
                          </mml:mtr>
                        </mml:mtable>
                      </mml:mrow>
                    </mml:math>
                  </disp-formula>
                </p>
                <p>where <italic>b<sub>1</sub></italic>, <italic>b<sub>2</sub></italic>, <italic>b<sub>3</sub></italic> represent the category reliability for the three types of sea ice, respectively.</p>
              </list-item>
            </list>
          </p>
          <boxed-text id="algorithm1">
            <list>
              <list-item>
                <p><bold>Input:</bold> The bounding box evidence <inline-formula><mml:math alttext="a_{i}" display="inline"><mml:msub><mml:mi>a</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:math></inline-formula>, The category evidence <inline-formula><mml:math alttext="b_{i}" display="inline"><mml:msub><mml:mi>b</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:math></inline-formula>;</p>
              </list-item>
              <list-item>
                <p><bold>Output:</bold> New category result <inline-formula><mml:math alttext="New\_cate_{i}" display="inline"><mml:mrow><mml:mi>N</mml:mi><mml:mo>⁢</mml:mo><mml:mi>e</mml:mi><mml:mo>⁢</mml:mo><mml:mi>w</mml:mi><mml:mo>⁢</mml:mo><mml:mi mathvariant="normal">_</mml:mi><mml:mo>⁢</mml:mo><mml:mi>c</mml:mi><mml:mo>⁢</mml:mo><mml:mi>a</mml:mi><mml:mo>⁢</mml:mo><mml:mi>t</mml:mi><mml:mo>⁢</mml:mo><mml:msub><mml:mi>e</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>;</p>
              </list-item>
              <list-item>
                <p/>
              </list-item>
              <list-item>
                <p><bold>for</bold> <italic><inline-formula><mml:math alttext="i=1,2,3" display="inline"><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mrow><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mn>3</mml:mn></mml:mrow></mml:mrow></mml:math></inline-formula></italic> <bold>do</bold></p>
              </list-item>
              <list-item>
                <p>   <inline-formula><mml:math alttext="ma_{i}=1-b_{i}" display="inline"><mml:mrow><mml:mrow><mml:mi>m</mml:mi><mml:mo>⁢</mml:mo><mml:msub><mml:mi>a</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mo>=</mml:mo><mml:mrow><mml:mn>1</mml:mn><mml:mo>−</mml:mo><mml:msub><mml:mi>b</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:mrow></mml:math></inline-formula>;</p>
              </list-item>
              <list-item>
                <p>   <inline-formula><mml:math alttext="mb_{i}=1-a_{i}" display="inline"><mml:mrow><mml:mrow><mml:mi>m</mml:mi><mml:mo>⁢</mml:mo><mml:msub><mml:mi>b</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mo>=</mml:mo><mml:mrow><mml:mn>1</mml:mn><mml:mo>−</mml:mo><mml:msub><mml:mi>a</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:mrow></mml:math></inline-formula>;</p>
              </list-item>
              <list-item>
                <p>   <inline-formula><mml:math alttext="B_{i}=a_{i}^{2}\cdot b_{i}+{\displaystyle\frac{a_{i}^{2}\cdot ma_{i}}{a_{i}+ma%&#10;_{i}}}+{\displaystyle\frac{a_{i}\cdot b_{i}^{2}\cdot mb_{i}}{b_{i}+mb_{i}}}" display="inline"><mml:mrow><mml:msub><mml:mi>B</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mrow><mml:mrow><mml:msubsup><mml:mi>a</mml:mi><mml:mi>i</mml:mi><mml:mn>2</mml:mn></mml:msubsup><mml:mo lspace="0.222em" rspace="0.222em">⋅</mml:mo><mml:msub><mml:mi>b</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac><mml:mrow><mml:mrow><mml:msubsup><mml:mi>a</mml:mi><mml:mi>i</mml:mi><mml:mn>2</mml:mn></mml:msubsup><mml:mo lspace="0.222em" rspace="0.222em">⋅</mml:mo><mml:mi>m</mml:mi></mml:mrow><mml:mo>⁢</mml:mo><mml:msub><mml:mi>a</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:msub><mml:mi>a</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mrow><mml:mi>m</mml:mi><mml:mo>⁢</mml:mo><mml:msub><mml:mi>a</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac><mml:mrow><mml:mrow><mml:msub><mml:mi>a</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo lspace="0.222em" rspace="0.222em">⋅</mml:mo><mml:msubsup><mml:mi>b</mml:mi><mml:mi>i</mml:mi><mml:mn>2</mml:mn></mml:msubsup><mml:mo lspace="0.222em" rspace="0.222em">⋅</mml:mo><mml:mi>m</mml:mi></mml:mrow><mml:mo>⁢</mml:mo><mml:msub><mml:mi>b</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:msub><mml:mi>b</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mrow><mml:mi>m</mml:mi><mml:mo>⁢</mml:mo><mml:msub><mml:mi>b</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:mrow></mml:mfrac></mml:mstyle></mml:mrow></mml:mrow></mml:math></inline-formula>;</p>
              </list-item>
              <list-item>
                <p>   </p>
              </list-item>
              <list-item>
                <p> end for</p>
              </list-item>
              <list-item>
                <p><inline-formula><mml:math alttext="sumB=\sum_{i=1}^{3}B_{i}" display="inline"><mml:mrow><mml:mrow><mml:mi>s</mml:mi><mml:mo>⁢</mml:mo><mml:mi>u</mml:mi><mml:mo>⁢</mml:mo><mml:mi>m</mml:mi><mml:mo>⁢</mml:mo><mml:mi>B</mml:mi></mml:mrow><mml:mo rspace="0.111em">=</mml:mo><mml:mrow><mml:msubsup><mml:mo>∑</mml:mo><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mn>3</mml:mn></mml:msubsup><mml:msub><mml:mi>B</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:mrow></mml:math></inline-formula>;</p>
              </list-item>
              <list-item>
                <p><inline-formula><mml:math alttext="New\_cate_{i}=0" display="inline"><mml:mrow><mml:mrow><mml:mi>N</mml:mi><mml:mo>⁢</mml:mo><mml:mi>e</mml:mi><mml:mo>⁢</mml:mo><mml:mi>w</mml:mi><mml:mo>⁢</mml:mo><mml:mi mathvariant="normal">_</mml:mi><mml:mo>⁢</mml:mo><mml:mi>c</mml:mi><mml:mo>⁢</mml:mo><mml:mi>a</mml:mi><mml:mo>⁢</mml:mo><mml:mi>t</mml:mi><mml:mo>⁢</mml:mo><mml:msub><mml:mi>e</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mo>=</mml:mo><mml:mn>0</mml:mn></mml:mrow></mml:math></inline-formula>;</p>
              </list-item>
              <list-item>
                <p><bold>for</bold> <italic><inline-formula><mml:math alttext="i=1,2,3" display="inline"><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mrow><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mn>3</mml:mn></mml:mrow></mml:mrow></mml:math></inline-formula></italic> <bold>do</bold></p>
              </list-item>
              <list-item>
                <p>   <inline-formula><mml:math alttext="New\_cate_{i}={\displaystyle\frac{B_{i}}{sumB}}" display="inline"><mml:mrow><mml:mrow><mml:mi>N</mml:mi><mml:mo>⁢</mml:mo><mml:mi>e</mml:mi><mml:mo>⁢</mml:mo><mml:mi>w</mml:mi><mml:mo>⁢</mml:mo><mml:mi mathvariant="normal">_</mml:mi><mml:mo>⁢</mml:mo><mml:mi>c</mml:mi><mml:mo>⁢</mml:mo><mml:mi>a</mml:mi><mml:mo>⁢</mml:mo><mml:mi>t</mml:mi><mml:mo>⁢</mml:mo><mml:msub><mml:mi>e</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac><mml:msub><mml:mi>B</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mrow><mml:mi>s</mml:mi><mml:mo>⁢</mml:mo><mml:mi>u</mml:mi><mml:mo>⁢</mml:mo><mml:mi>m</mml:mi><mml:mo>⁢</mml:mo><mml:mi>B</mml:mi></mml:mrow></mml:mfrac></mml:mstyle></mml:mrow></mml:math></inline-formula>;</p>
              </list-item>
              <list-item>
                <p>   </p>
              </list-item>
              <list-item>
                <p> end for</p>
              </list-item>
              <list-item>
                <p/>
              </list-item>
            </list>
            <label>Algorithm 1</label>
            <caption>
              <p>Optimized DSmT fusion inference algorithm</p>
            </caption>
          </boxed-text>
          <p>
            <fig-group>
              <fig id="F11.a">
                <caption>
                  <p>(a) Small-scale sea ice: the longest side of the circumscribed rectangle for this type of sea ice ranges between 8 pixels and 32 pixels.</p>
                  <graphic xlink:href="fig12(a)small.pdf"/>
                </caption>
              </fig>
              <fig id="F.b">
                <caption>
                  <p>(b) Medium-scale sea ice: the longest side of the circumscribed rectangle for this type of sea ice ranges between 32 pixels and 128 pixels.</p>
                  <graphic xlink:href="fig12(b)medium.pdf"/>
                </caption>
              </fig>
              <fig id="F.c">
                <caption>
                  <p>(c) Large-scale sea ice: the longest side of the circumscribed rectangle for this type of sea ice exceeds 128 pixels.</p>
                  <graphic xlink:href="fig12(c)big.pdf"/>
                </caption>
              </fig>
              <fig id="F11">
                <label>Figure 11.</label>
                <caption>
                  <p>Three distinct categories of sea ice.</p>
                </caption>
              </fig>
            </fig-group>
          </p>
          <p>
            <table-wrap id="T1">
              <label>Table 1</label>
              <caption>
                <p>Summary of the labels for the three types of sea ice.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;"/>
                  <td style="border-top: 1px solid black;"/>
                  <td style="border-top: 1px solid black;" align="center">Small-Scale Sea Ice</td>
                  <td style="border-top: 1px solid black;" align="center">Medium-Scale Sea Ice</td>
                  <td style="border-top: 1px solid black;" align="center">Big-Scale Sea Ice</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">NWPU- RESISC45 [<xref rid="ref029" ref-type="bibr">29</xref>]</td>
                  <td style="border-top: 1px solid black;" align="center">Quantity</td>
                  <td style="border-top: 1px solid black;" align="center">6820</td>
                  <td style="border-top: 1px solid black;" align="center">3710</td>
                  <td style="border-top: 1px solid black;" align="center">382</td>
                </tr>
                <tr>
                  <td align="center">Percentage</td>
                  <td align="center">62.5%</td>
                  <td align="center">34.0%</td>
                  <td align="center">3.5%</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">Our Sea Ice Dataset</td>
                  <td style="border-top: 1px solid black;" align="center">Quantity</td>
                  <td style="border-top: 1px solid black;" align="center">3162</td>
                  <td style="border-top: 1px solid black;" align="center">1170</td>
                  <td style="border-top: 1px solid black;" align="center">256</td>
                </tr>
                <tr>
                  <td align="center">Percentage</td>
                  <td align="center">69.0%</td>
                  <td align="center">25.5%</td>
                  <td align="center">5.5%</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" colspan="5" align="left">
                    <p>Note: We randomly select 70% of the images to train the model, and the remaining 30% of the images were used to verify the training effect.</p>
                  </td>
                </tr>
              </table>
            </table-wrap>
          </p>
          <p>
            <table-wrap id="T2">
              <label>Table 2</label>
              <caption>
                <p>Important information about the exclusive Landsat8-based sea ice dataset.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">Attribute</td>
                  <td style="border-top: 1px solid black;" colspan="2" align="center">Attribute Value</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" align="center">1</td>
                  <td style="border-top: 1px solid black;" align="center">2</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" align="center">SPACECRAFT_ID</td>
                  <td style="border-top: 1px solid black;" align="center">LANDSAT8</td>
                  <td style="border-top: 1px solid black;" align="center">LANDSAT8</td>
                </tr>
                <tr>
                  <td align="center">ORIGIN</td>
                  <td align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">Image courtesy of</td>
                          </tr>
                          <tr>
                            <td align="center">the U.S. Geological Survey</td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">Image courtesy of</td>
                          </tr>
                          <tr>
                            <td align="center">the U.S. Geological Survey</td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                </tr>
                <tr>
                  <td align="center">LANDSAT_SCENE_ID</td>
                  <td align="center">LC80482392019215LGN00</td>
                  <td align="center">LC81300082018179LGN00</td>
                </tr>
                <tr>
                  <td align="center">LANDSAT_PRODUCT_ID</td>
                  <td align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">LC08_L1GT_048239_</td>
                          </tr>
                          <tr>
                            <td align="center">20190803_20190819_01_T2</td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">LC08_L1TP_130008_</td>
                          </tr>
                          <tr>
                            <td align="center">20180628_20180704_01_T1</td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                </tr>
                <tr>
                  <td align="center">FILE_DATE</td>
                  <td align="center">2019-08-19T23:27:47Z</td>
                  <td align="center">2018-07-04T09:14:55Z</td>
                </tr>
                <tr>
                  <td align="center">OUTPUT_FORMAT</td>
                  <td align="center">GEOTIFF</td>
                  <td align="center">GEOTIFF</td>
                </tr>
                <tr>
                  <td align="center">SENSOR_ID</td>
                  <td align="center">OLI_TIRS</td>
                  <td align="center">OLI_TIRS</td>
                </tr>
                <tr>
                  <td align="center">TARGET_WRS_PATH</td>
                  <td align="center">48</td>
                  <td align="center">130</td>
                </tr>
                <tr>
                  <td align="center">TARGET_WRS_ROW</td>
                  <td align="center">239</td>
                  <td align="center">8</td>
                </tr>
                <tr>
                  <td align="center">DATE_ACQUIRED</td>
                  <td align="center">2019-08-03</td>
                  <td align="center">2018-06-28</td>
                </tr>
                <tr>
                  <td align="center">SCENE_CENTER_TIME</td>
                  <td align="center">20:32:39.0212890Z</td>
                  <td align="center">03:26:15.2127540Z</td>
                </tr>
                <tr>
                  <td align="center">CLOUD_COVER</td>
                  <td align="center">1.55</td>
                  <td align="center">2.20</td>
                </tr>
                <tr>
                  <td align="center">CLOUD_COVER_LAND</td>
                  <td align="center">0.02</td>
                  <td align="center">0.12</td>
                </tr>
                <tr>
                  <td align="center">IMAGE_QUALITY_OLI</td>
                  <td align="center">9</td>
                  <td align="center">9</td>
                </tr>
                <tr>
                  <td align="center">IMAGE_QUALITY_TIRS</td>
                  <td align="center">9</td>
                  <td align="center">9</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" colspan="3" align="left">
                    <p>Note: This dataset pertains to the sea ice data corresponding to the aforementioned two scenes. The dataset comprises a total of 430 images, each with dimensions of 256 * 256 pixels.</p>
                  </td>
                </tr>
              </table>
            </table-wrap>
          </p>
          <p>
            <table-wrap id="T3">
              <label>Table 3</label>
              <caption>
                <p>Experimental configuration.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;" align="center">Attribute</td>
                  <td style="border-top: 1px solid black;" align="center">Attribute Value</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" align="center">CPU</td>
                  <td style="border-top: 1px solid black;" align="center">Core i5 12450H</td>
                </tr>
                <tr>
                  <td align="center">GPU</td>
                  <td align="center">NVIDIA GeForce RTX 3050</td>
                </tr>
                <tr>
                  <td align="center">Running memory</td>
                  <td align="center">16GB</td>
                </tr>
                <tr>
                  <td align="center">Storage memory</td>
                  <td align="center">256GB</td>
                </tr>
                <tr>
                  <td align="center">Operating system</td>
                  <td align="center">Win 10</td>
                </tr>
                <tr>
                  <td align="center">Interpreter</td>
                  <td align="center">Python 3.9</td>
                </tr>
                <tr>
                  <td align="center">Deep Learning Frameworks</td>
                  <td align="center">PyTorch 1.9</td>
                </tr>
                <tr>
                  <td style="border-bottom: 1px solid black;" align="center">IDEA</td>
                  <td style="border-bottom: 1px solid black;" align="center">PyCharm</td>
                </tr>
              </table>
            </table-wrap>
          </p>
          <p>
            <table-wrap id="T4">
              <label>Table 4</label>
              <caption>
                <p>Hyper-parameters of improved YOLOv8 Algorithm.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;" align="center">Attribute</td>
                  <td style="border-top: 1px solid black;" align="center">Attribute Value</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" align="center">epochs</td>
                  <td style="border-top: 1px solid black;" align="center">500</td>
                </tr>
                <tr>
                  <td align="center">batch size</td>
                  <td align="center">16</td>
                </tr>
                <tr>
                  <td align="center">imgsz</td>
                  <td align="center">256</td>
                </tr>
                <tr>
                  <td align="center">workers</td>
                  <td align="center">8</td>
                </tr>
                <tr>
                  <td align="center">close mosaic</td>
                  <td align="center">Last 10 epochs</td>
                </tr>
                <tr>
                  <td align="center">optimizer</td>
                  <td align="center">AdamW</td>
                </tr>
                <tr>
                  <td align="center">initial learning rate</td>
                  <td align="center">0.01</td>
                </tr>
                <tr>
                  <td align="center">final learning rate</td>
                  <td align="center">0.0001</td>
                </tr>
                <tr>
                  <td align="center">momentum</td>
                  <td align="center">0.937</td>
                </tr>
                <tr>
                  <td align="center">weight decay</td>
                  <td align="center">0.0005</td>
                </tr>
                <tr>
                  <td align="center">warm-up epochs</td>
                  <td align="center">3.0</td>
                </tr>
                <tr>
                  <td align="center">warm-up momentum</td>
                  <td align="center">0.8</td>
                </tr>
                <tr>
                  <td align="center">warm-up bias learning rate</td>
                  <td align="center">0.1</td>
                </tr>
                <tr>
                  <td align="center">box loss gain</td>
                  <td align="center">7.5</td>
                </tr>
                <tr>
                  <td align="center">class loss gain</td>
                  <td align="center">0.5</td>
                </tr>
                <tr>
                  <td align="center">DFL loss gain</td>
                  <td align="center">1.5</td>
                </tr>
                <tr>
                  <td align="center">hsv hue augmentation</td>
                  <td align="center">0.015</td>
                </tr>
                <tr>
                  <td align="center">hsv saturation augmentation</td>
                  <td align="center">0.7</td>
                </tr>
                <tr>
                  <td align="center">hsv value augmentation</td>
                  <td align="center">0.4</td>
                </tr>
                <tr>
                  <td align="center">translation augmentation</td>
                  <td align="center">0.1</td>
                </tr>
                <tr>
                  <td align="center">scale augmentation</td>
                  <td align="center">0.9</td>
                </tr>
                <tr>
                  <td align="center">mosaic augmentation</td>
                  <td align="center">1.0</td>
                </tr>
                <tr>
                  <td align="center">mixup augmentation</td>
                  <td align="center">0.1</td>
                </tr>
                <tr>
                  <td style="border-bottom: 1px solid black;" align="center">copy-paste augmentation</td>
                  <td style="border-bottom: 1px solid black;" align="center">0.1</td>
                </tr>
              </table>
            </table-wrap>
          </p>
          <p>
            <table-wrap id="T5">
              <label>Table 5</label>
              <caption>
                <p>Comparisons with the baseline model and state-of-the-arts.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;" align="left">Model</td>
                  <td style="border-top: 1px solid black;" align="center">Precision (%)</td>
                  <td style="border-top: 1px solid black;" align="center">Recall (%)</td>
                  <td style="border-top: 1px solid black;" align="center">mAP50 (%)</td>
                  <td style="border-top: 1px solid black;" align="center">mAP50-95 (%)</td>
                  <td style="border-top: 1px solid black;" align="center">F1 (%)</td>
                  <td style="border-top: 1px solid black;" align="center">Training time (h)</td>
                  <td style="border-top: 1px solid black;" align="center">FPS</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" align="left">Faster R-CNN</td>
                  <td style="border-top: 1px solid black;" align="center">62.7</td>
                  <td style="border-top: 1px solid black;" align="center">80.3</td>
                  <td style="border-top: 1px solid black;" align="center">79.2</td>
                  <td style="border-top: 1px solid black;" align="center">47.1</td>
                  <td style="border-top: 1px solid black;" align="center">70.4</td>
                  <td style="border-top: 1px solid black;" align="center">6.09</td>
                  <td style="border-top: 1px solid black;" align="center">4.4</td>
                </tr>
                <tr>
                  <td align="left">SSD</td>
                  <td align="center">62.3</td>
                  <td align="center">76.0</td>
                  <td align="center">78.0</td>
                  <td align="center">47.6</td>
                  <td align="center">68.5</td>
                  <td align="center">0.98</td>
                  <td align="center">6.9</td>
                </tr>
                <tr>
                  <td align="left">RT-DETR</td>
                  <td align="center">74.9</td>
                  <td align="center">66.1</td>
                  <td align="center">73</td>
                  <td align="center">54.2</td>
                  <td align="center">70.2</td>
                  <td align="center">3.213</td>
                  <td align="center">68.0</td>
                </tr>
                <tr>
                  <td align="left">YOLOv3</td>
                  <td align="center">66.8</td>
                  <td align="center">86</td>
                  <td align="center">74.7</td>
                  <td align="center">50.9</td>
                  <td align="center">75.2</td>
                  <td align="center">5.719</td>
                  <td align="center">59.2</td>
                </tr>
                <tr>
                  <td align="left">YOLOv5</td>
                  <td align="center">73.3</td>
                  <td align="center">73.3</td>
                  <td align="center">82.2</td>
                  <td align="center">52.5</td>
                  <td align="center">73.3</td>
                  <td align="center">0.728</td>
                  <td align="center">
                    <bold>208.3</bold>
                  </td>
                </tr>
                <tr>
                  <td align="left">YOLOv6</td>
                  <td align="center">72.4</td>
                  <td align="center">69.3</td>
                  <td align="center">78.4</td>
                  <td align="center">47.0</td>
                  <td align="center">70.8</td>
                  <td align="center">3.262</td>
                  <td align="center">69.4</td>
                </tr>
                <tr>
                  <td align="left">YOLOv7</td>
                  <td align="center">76.7</td>
                  <td align="center">64.6</td>
                  <td align="center">81.5</td>
                  <td align="center">49.1</td>
                  <td align="center">70.1</td>
                  <td align="center">4.529</td>
                  <td align="center">87.7</td>
                </tr>
                <tr>
                  <td align="left">YOLOv9</td>
                  <td align="center">65.6</td>
                  <td align="center">
                    <bold>82.3</bold>
                  </td>
                  <td align="center">78.0</td>
                  <td align="center">44.2</td>
                  <td align="center">73.0</td>
                  <td align="center">5.46</td>
                  <td align="center">108.7</td>
                </tr>
                <tr>
                  <td align="left">YOLOv10</td>
                  <td align="center">84.2</td>
                  <td align="center">68.1</td>
                  <td align="center">81.9</td>
                  <td align="center">49.0</td>
                  <td align="center">75.3</td>
                  <td align="center">
                    <bold>0.483</bold>
                  </td>
                  <td align="center">128.2</td>
                </tr>
                <tr>
                  <td align="left">ASF-YOLO</td>
                  <td align="center">64.6</td>
                  <td align="center">77.9</td>
                  <td align="center">80.3</td>
                  <td align="center">45.1</td>
                  <td align="center">70.6</td>
                  <td align="center">3.983</td>
                  <td align="center">53.5</td>
                </tr>
                <tr>
                  <td align="left">GOLD-YOLO</td>
                  <td align="center">73.4</td>
                  <td align="center">74.7</td>
                  <td align="center">81.1</td>
                  <td align="center">47.8</td>
                  <td align="center">74.0</td>
                  <td align="center">6.118</td>
                  <td align="center">58.5</td>
                </tr>
                <tr>
                  <td align="left">Hyper-YOLO</td>
                  <td align="center">69.2</td>
                  <td align="center">76.9</td>
                  <td align="center">83.1</td>
                  <td align="center">50.5</td>
                  <td align="center">72.8</td>
                  <td align="center">6.307</td>
                  <td align="center">44.4</td>
                </tr>
                <tr>
                  <td align="left">Improved YOLOv5[<xref rid="ref039" ref-type="bibr">39</xref>]</td>
                  <td align="center">71.2</td>
                  <td align="center">75.1</td>
                  <td align="center">82.6</td>
                  <td align="center">51.5</td>
                  <td align="center">73.1</td>
                  <td align="center">3.52</td>
                  <td align="center">126.4</td>
                </tr>
                <tr>
                  <td align="left">YOLOv8</td>
                  <td align="center">
                    <bold>84.7</bold>
                  </td>
                  <td align="center">62.4</td>
                  <td align="center">81.6</td>
                  <td align="center">56.2</td>
                  <td align="center">71.9</td>
                  <td align="center">1.113</td>
                  <td align="center">82.6</td>
                </tr>
                <tr>
                  <td align="left">
                    <bold>Our Improved YOLOv8</bold>
                  </td>
                  <td align="center">79.4</td>
                  <td align="center">78.0</td>
                  <td align="center">
                    <bold>87.2</bold>
                  </td>
                  <td align="center">
                    <bold>59.3</bold>
                  </td>
                  <td align="center">
                    <bold>78.7</bold>
                  </td>
                  <td align="center">0.959</td>
                  <td align="center">48.3</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" colspan="8" align="left">In the same group of experiments, the best-performing data is highlighted in bold.</td>
                </tr>
              </table>
            </table-wrap>
          </p>
          <p>
            <fig id="F12">
              <label>Figure 12.</label>
              <caption>
                <p>Schematic diagram of the sea ice satellite imagery.</p>
              </caption>
              <graphic xlink:href="fig11map.pdf"/>
            </fig>
          </p>
        </sec>
      </sec>
    </sec>
    <sec id="S4">
      <label>4.</label>
      <title>Experiments</title>
      <sec id="S4.SS1">
        <label>4.1</label>
        <title>Data Collection</title>
        <p id="S4.SS1.p1">We currently employ two distinct sea ice datasets to assess the detection accuracy of our improved YOLOv8 model. The first dataset is the widely recognized NWPU-RESISC45 [<xref rid="ref029" ref-type="bibr">29</xref>], while the second consists of an exclusive Landsat8-based sea ice dataset that we have developed.</p>
        <sec id="S4.SS1.SSS1">
          <label>4.1.1</label>
          <title>A Sea Ice Dataset Derived from NWPU-RESISC45 </title>
          <p id="S4.SS1.SSS1.p1">When traversing sea-ice laden waters, it is imperative for the crew to swiftly discern and precisely locate sea ice of diverse dimensions to enable the vessel to bypass the perilous sea ice. Consequently, we categorize the sea ice in these two datasets into three distinct categories. We employ the software labelimg [<xref rid="ref055" ref-type="bibr">55</xref>] to annotate the images of three distinct types of sea ice, designated as <italic>Small-scale sea ice</italic>, <italic>Medium-scale sea ice</italic>, and <italic>Large-scale sea ice</italic>, as shown in Figure <xref ref-type="fig" rid="F10">10</xref>.</p>
        </sec>
        <sec id="S4.SS1.SSS2">
          <label>4.1.2</label>
          <title>An Landsat8-based Sea Ice Dataset</title>
          <p id="S4.SS1.SSS2.p1">We cropped the satellite data into uniformly sized images and utilized the software labelimg [<xref rid="ref055" ref-type="bibr">55</xref>] to annotate three distinct types of sea ice present in these images, designated as <italic>Small-scale sea ice</italic>, <italic>Medium-scale sea ice</italic>, and <italic>Large-scale sea ice</italic>, as shown in Figure <xref ref-type="fig" rid="F11">11</xref>. The specifics of this dataset are provided in the appendix, as shown in Tables <xref rid="T1" ref-type="table">1</xref> and <xref rid="T2" ref-type="table">2</xref>.</p>
          <p id="S4.SS1.SSS2.p2">In addition to the NWPU-RESISC45 [<xref rid="ref029" ref-type="bibr">29</xref>] dataset, we also explored other datasets to continually validate the performance of our enhanced YOLOv8-based sea ice detector. We identified areas where sea ice occurs at high latitudes and acquired satellite data for these regions, as depicted in Figure <xref ref-type="fig" rid="F12">12</xref>.</p>
        </sec>
      </sec>
      <sec id="S4.SS2">
        <label>4.2</label>
        <title>Implementation Details</title>
        <p>
          <table-wrap id="T6">
            <label>Table 6</label>
            <caption>
              <p>Comparisons with the baseline model and state-of-the-arts.</p>
            </caption>
            <table>
              <tr>
                <td style="border-top: 1px solid black;" align="left">Model</td>
                <td style="border-top: 1px solid black;" align="center">Precision (%)</td>
                <td style="border-top: 1px solid black;" align="center">Recall (%)</td>
                <td style="border-top: 1px solid black;" align="center">mAP50 (%)</td>
                <td style="border-top: 1px solid black;" align="center">mAP50-95 (%)</td>
                <td style="border-top: 1px solid black;" align="center">F1 (%)</td>
                <td style="border-top: 1px solid black;" align="center">Training time (h)</td>
                <td style="border-top: 1px solid black;" align="center">FPS</td>
              </tr>
              <tr>
                <td style="border-top: 1px solid black;" align="left">Faster-RCNN</td>
                <td style="border-top: 1px solid black;" align="center">70.7</td>
                <td style="border-top: 1px solid black;" align="center">84.1</td>
                <td style="border-top: 1px solid black;" align="center">83.2</td>
                <td style="border-top: 1px solid black;" align="center">66.2</td>
                <td style="border-top: 1px solid black;" align="center">76.8</td>
                <td style="border-top: 1px solid black;" align="center">14.79</td>
                <td style="border-top: 1px solid black;" align="center">5.8</td>
              </tr>
              <tr>
                <td align="left">SSD</td>
                <td align="center">72.1</td>
                <td align="center">71.4</td>
                <td align="center">72.0</td>
                <td align="center">53.5</td>
                <td align="center">71.7</td>
                <td align="center">2.38</td>
                <td align="center">6.9</td>
              </tr>
              <tr>
                <td align="left">RT-DETR</td>
                <td align="center">66.7</td>
                <td align="center">67.2</td>
                <td align="center">73.6</td>
                <td align="center">45.4</td>
                <td align="center">66.9</td>
                <td align="center">7.871</td>
                <td align="center">40.5</td>
              </tr>
              <tr>
                <td align="left">YOLOv3</td>
                <td align="center">73.1</td>
                <td align="center">71.3</td>
                <td align="center">81.4</td>
                <td align="center">50.2</td>
                <td align="center">72.2</td>
                <td align="center">6.29</td>
                <td align="center">53.8</td>
              </tr>
              <tr>
                <td align="left">YOLOv5</td>
                <td align="center">72.5</td>
                <td align="center">88.9</td>
                <td align="center">90.2</td>
                <td align="center">69.5</td>
                <td align="center">79.9</td>
                <td align="center">2.499</td>
                <td align="center">156.4</td>
              </tr>
              <tr>
                <td align="left">YOLOv6</td>
                <td align="center">74.0</td>
                <td align="center">77.0</td>
                <td align="center">80.4</td>
                <td align="center">46.1</td>
                <td align="center">75.5</td>
                <td align="center">13.668</td>
                <td align="center">52.8</td>
              </tr>
              <tr>
                <td align="left">YOLOv7</td>
                <td align="center">74.9</td>
                <td align="center">74.6</td>
                <td align="center">82.8</td>
                <td align="center">53.0</td>
                <td align="center">74.7</td>
                <td align="center">2.142</td>
                <td align="center">
                  <bold>192.3</bold>
                </td>
              </tr>
              <tr>
                <td align="left">YOLOv9</td>
                <td align="center">86.8</td>
                <td align="center">70.3</td>
                <td align="center">84.7</td>
                <td align="center">52.8</td>
                <td align="center">77.7</td>
                <td align="center">1.921</td>
                <td align="center">181.8</td>
              </tr>
              <tr>
                <td align="left">YOLOv10</td>
                <td align="center">77.5</td>
                <td align="center">
                  <bold>91.6</bold>
                </td>
                <td align="center">91.3</td>
                <td align="center">62.2</td>
                <td align="center">84.0</td>
                <td align="center">
                  <bold>1.581</bold>
                </td>
                <td align="center">112.8</td>
              </tr>
              <tr>
                <td align="left">ASF-YOLO</td>
                <td align="center">77.3</td>
                <td align="center">90.0</td>
                <td align="center">91.6</td>
                <td align="center">68.5</td>
                <td align="center">83.2</td>
                <td align="center">5.865</td>
                <td align="center">61.7</td>
              </tr>
              <tr>
                <td align="left">GOLD-YOLO</td>
                <td align="center">81.7</td>
                <td align="center">65.9</td>
                <td align="center">82.1</td>
                <td align="center">51.6</td>
                <td align="center">73.0</td>
                <td align="center">6.919</td>
                <td align="center">55.6</td>
              </tr>
              <tr>
                <td align="left">Hyper-YOLO</td>
                <td align="center">86.2</td>
                <td align="center">66.4</td>
                <td align="center">86.3</td>
                <td align="center">55.0</td>
                <td align="center">75.0</td>
                <td align="center">8.5</td>
                <td align="center">70.9</td>
              </tr>
              <tr>
                <td align="left">Improved YOLOv5[<xref rid="ref039" ref-type="bibr">39</xref>]</td>
                <td align="center">75.1</td>
                <td align="center">90.3</td>
                <td align="center">90.1</td>
                <td align="center">65.9</td>
                <td align="center">82.0</td>
                <td align="center">2.042</td>
                <td align="center">134.7</td>
              </tr>
              <tr>
                <td align="left">YOLOv8</td>
                <td align="center">86.1</td>
                <td align="center">85.6</td>
                <td align="center">92.7</td>
                <td align="center">67.6</td>
                <td align="center">85.8</td>
                <td align="center">9.435</td>
                <td align="center">50.8</td>
              </tr>
              <tr>
                <td align="left">
                  <bold>Our Improved YOLOv8</bold>
                </td>
                <td align="center">
                  <bold>90.6</bold>
                </td>
                <td align="center">78.0</td>
                <td align="center">
                  <bold>93.8</bold>
                </td>
                <td align="center">
                  <bold>71.8</bold>
                </td>
                <td align="center">
                  <bold>87.4</bold>
                </td>
                <td align="center">4.624</td>
                <td align="center">53.2</td>
              </tr>
              <tr>
                <td style="border-top: 1px solid black;" colspan="8" align="left">In the same group of experiments, the best-performing data is highlighted in bold.</td>
              </tr>
            </table>
          </table-wrap>
        </p>
        <p id="S4.SS2.p1">We use YOLOv8 as a baseline model. Since the release of YOLOv8 in 2023, it has been deployed on various types of hardware due to its low resource requirements. The following are the experimental configuration, hyper-parameters of improved YOLOv8 object detector and summary of the labels for the three types of sea ice, as shown in Tables <xref rid="T3" ref-type="table">3</xref> and <xref rid="T4" ref-type="table">4</xref>.</p>
        <p>
          <fig-group>
            <fig id="F13.a">
              <caption>
                <p>(a) YOLOv8 incorrectly predicts medium-scale sea ice as large-scale sea ice.</p>
                <graphic xlink:href="fig13a.pdf"/>
              </caption>
            </fig>
            <fig id="F13.b">
              <caption>
                <p>(b) Our improved YOLOv8 correctly predicts the results.</p>
                <graphic xlink:href="fig13b.pdf"/>
              </caption>
            </fig>
            <fig id="F13.c">
              <caption>
                <p>(c) YOLOv8 incorrectly predicts large-scale sea ice as medium-scale sea ice.</p>
                <graphic xlink:href="fig13c.pdf"/>
              </caption>
            </fig>
            <fig id="F13.d">
              <caption>
                <p>(d) Our improved YOLOv8 correctly predicts the results.</p>
                <graphic xlink:href="fig13d.pdf"/>
              </caption>
            </fig>
              <fig id="F13.e">
                <caption>
                  <p>(e) YOLOv8 incorrectly predicts medium-scale sea ice as small-scale sea ice.</p>
                  <graphic xlink:href="fig13e.pdf"/>
                </caption>
              </fig>
              <fig id="F13.f">
                <caption>
                  <p>(f) Our improved YOLOv8 correctly predicts the results.</p>
                  <graphic xlink:href="fig13f.pdf"/>
                </caption>
              </fig>
              <fig id="F13.g">
                <caption>
                  <p>(g) YOLOv8 outputs redundant prediction boxes.</p>
                  <graphic xlink:href="fig13g.pdf"/>
                </caption>
              </fig>
              <fig id="F13.h">
                <caption>
                  <p>(h) Our improved YOLOv8 correctly predicts the results.</p>
                  <graphic xlink:href="fig13h.pdf"/>
                </caption>
              </fig>
              <fig id="F13">
                <label>Figure 13.</label>
                <caption>
                  <p>Detection results comparison between YOLOv8 and our improved model.</p>
                </caption>
              </fig>
          </fig-group>
        </p>
      </sec>
      <sec id="S4.SS3">
        <label>4.3</label>
        <title>Comparison with State-of-the-Arts</title>
        <sec id="S4.SS3.SSS1">
          <label>4.3.1</label>
          <title>Experimental Results Utilizing the NWPU-RESISC45 Dataset</title>
          <p id="S4.SS3.SSS1.p1">As shown in Table <xref rid="T5" ref-type="table">5</xref>, we conduct experiments on the NWPU-RESISC45 Dataset with mainstream object detection algorithms and perform a comparative analysis of our improved YOLOv8. The selected object detection algorithms include: the two-stage object detection algorithm Faster R-CNN, the Transformer-based object detection algorithm RT-DETR, the one-stage object detection algorithms from the YOLO series, and the improved YOLO series algorithms. Compared to YOLOv8, our improved YOLOv8 achieves a 15.6% increase in Recall, a 5.6% improvement in mAP50, a 3.1% enhancement in mAP50-95, and a 6.8% boost in F1 score, while simultaneously reducing the training time by 13.8%.</p>
          <p id="S4.SS3.SSS1.p2">In addition, compared to other improved YOLO series algorithms, our enhanced YOLOv8 also demonstrates outstanding detection accuracy and relatively faster convergence speed. As Figure <xref ref-type="fig" rid="F13">13</xref> illustrates, we present the detection effects of our improved YOLOv8 alongside the baseline model YOLOv8.</p>
          <p>
            <table-wrap id="T7">
              <label>Table 7</label>
              <caption>
                <p>Ablation study with improved YOLOv8.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;" align="justify">
                    <p>
                      <bold>Dataset</bold>
                    </p>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>Method</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>Fusion Module</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>Shape-IoU</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>Evidence Fusion</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>mAP50 (%)</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>mAP50-95 (%)</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="5" align="justify">
                    <p>NWPU-RESISC45 [<xref rid="ref029" ref-type="bibr">29</xref>]</p>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">YOLOv8</td>
                  <td style="border-top: 1px solid black;"/>
                  <td style="border-top: 1px solid black;"/>
                  <td style="border-top: 1px solid black;"/>
                  <td style="border-top: 1px solid black;" align="center">81.6</td>
                  <td style="border-top: 1px solid black;" align="center">56.2</td>
                </tr>
                <tr>
                  <td align="center">Algorithm 1</td>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td/>
                  <td/>
                  <td align="center">83.3</td>
                  <td align="center">57.0</td>
                </tr>
                <tr>
                  <td align="center">Algorithm 2</td>
                  <td/>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td/>
                  <td align="center">82.8</td>
                  <td align="center">58.3</td>
                </tr>
                <tr>
                  <td align="center">Algorithm 3</td>
                  <td/>
                  <td/>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td align="center">85.3</td>
                  <td align="center">56.8</td>
                </tr>
                <tr>
                  <td align="center">
                    <bold>Our Improved YOLOv8</bold>
                  </td>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td align="center">
                    <bold>87.2</bold>
                  </td>
                  <td align="center">
                    <bold>59.3</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="5" align="justify">
                    <p>Our Sea Ice Dataset</p>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">YOLOv8</td>
                  <td style="border-top: 1px solid black;"/>
                  <td style="border-top: 1px solid black;"/>
                  <td style="border-top: 1px solid black;"/>
                  <td style="border-top: 1px solid black;" align="center">92.7</td>
                  <td style="border-top: 1px solid black;" align="center">67.6</td>
                </tr>
                <tr>
                  <td align="center">Algorithm 1</td>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td/>
                  <td/>
                  <td align="center">92.9</td>
                  <td align="center">68.2</td>
                </tr>
                <tr>
                  <td align="center">Algorithm 2</td>
                  <td/>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td/>
                  <td align="center">93.2</td>
                  <td align="center">70.1</td>
                </tr>
                <tr>
                  <td align="center">Algorithm 3</td>
                  <td/>
                  <td/>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td align="center">93.3</td>
                  <td align="center">68.0</td>
                </tr>
                <tr>
                  <td align="center">
                    <bold>Our Improved YOLOv8</bold>
                  </td>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td align="center">
                    <inline-formula>
                      <mml:math alttext="\checkmark" display="inline">
                        <mml:mi mathvariant="normal">✓</mml:mi>
                      </mml:math>
                    </inline-formula>
                  </td>
                  <td align="center">
                    <bold>93.8</bold>
                  </td>
                  <td align="center">
                    <bold>71.8</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" colspan="7" align="justify">Note: <inline-formula><mml:math alttext="\checkmark" display="inline"><mml:mi mathvariant="normal">✓</mml:mi></mml:math></inline-formula> denotes an added module based on YOLOv8. In the same group of experiments, the best-performing data is highlighted in bold.</td>
                </tr>
              </table>
            </table-wrap>
          </p>
          <p>
            <table-wrap id="T8">
              <label>Table 8</label>
              <caption>
                <p>Ablation study with fusion module.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="2" align="left">
                    <bold>Dataset</bold>
                  </td>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">
                    <bold>Method</bold>
                  </td>
                  <td style="border-top: 1px solid black;" colspan="3" align="center">
                    <bold>AP50(%)</bold>
                  </td>
                  <td style="border-top: 1px solid black;" rowspan="2" align="left">
                    <bold>mAP50 (%)</bold>
                  </td>
                  <td style="border-top: 1px solid black;" rowspan="2" align="left">
                    <bold>mAP50-95 (%)</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>small</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>medium</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>big</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="6" align="left">NWPU-RESISC45 [<xref rid="ref029" ref-type="bibr">29</xref>]</td>
                  <td style="border-top: 1px solid black;" align="center">YOLOv8</td>
                  <td style="border-top: 1px solid black;" align="center">80.3</td>
                  <td style="border-top: 1px solid black;" align="center">80.8</td>
                  <td style="border-top: 1px solid black;" align="center">83.7</td>
                  <td style="border-top: 1px solid black;" align="center">81.6</td>
                  <td style="border-top: 1px solid black;" align="center">56.2</td>
                </tr>
                <tr>
                  <td align="center">+ SE</td>
                  <td align="center">52.3</td>
                  <td align="center">
                    <bold>87.2</bold>
                  </td>
                  <td align="center">
                    <bold>88.3</bold>
                  </td>
                  <td align="center">75.9</td>
                  <td align="center">51.7</td>
                </tr>
                <tr>
                  <td align="center">+ EMA</td>
                  <td align="center">69.0</td>
                  <td align="center">78.9</td>
                  <td align="center">81.3</td>
                  <td align="center">76.4</td>
                  <td align="center">48.0</td>
                </tr>
                <tr>
                  <td align="center">+ CA</td>
                  <td align="center">62.0</td>
                  <td align="center">74.7</td>
                  <td align="center">76.2</td>
                  <td align="center">71.0</td>
                  <td align="center">52.2</td>
                </tr>
                <tr>
                  <td align="center">+ CBAM</td>
                  <td align="center">79.5</td>
                  <td align="center">76.1</td>
                  <td align="center">83.9</td>
                  <td align="center">79.8</td>
                  <td align="center">53.5</td>
                </tr>
                <tr>
                  <td align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">
                              <bold>+ Fusion Module</bold>
                            </td>
                          </tr>
                          <tr>
                            <td align="center">
                              <bold>(Algorithm 1)</bold>
                            </td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td align="center">
                    <bold>84.5</bold>
                  </td>
                  <td align="center">81.0</td>
                  <td align="center">84.4</td>
                  <td align="center">
                    <bold>83.3</bold>
                  </td>
                  <td align="center">
                    <bold>57.0</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="6" align="left">Our Sea Ice Dataset</td>
                  <td style="border-top: 1px solid black;" align="center">YOLOv8</td>
                  <td style="border-top: 1px solid black;" align="center">89.1</td>
                  <td style="border-top: 1px solid black;" align="center">94.0</td>
                  <td style="border-top: 1px solid black;" align="center">95.0</td>
                  <td style="border-top: 1px solid black;" align="center">92.7</td>
                  <td style="border-top: 1px solid black;" align="center">67.6</td>
                </tr>
                <tr>
                  <td align="center">+ SE</td>
                  <td align="center">79.7</td>
                  <td align="center">91.8</td>
                  <td align="center">93.2</td>
                  <td align="center">88.2</td>
                  <td align="center">65.6</td>
                </tr>
                <tr>
                  <td align="center">+ EMA</td>
                  <td align="center">84.4</td>
                  <td align="center">91.1</td>
                  <td align="center">92.2</td>
                  <td align="center">89.3</td>
                  <td align="center">60.3</td>
                </tr>
                <tr>
                  <td align="center">+ CA</td>
                  <td align="center">85.9</td>
                  <td align="center">91.8</td>
                  <td align="center">94.3</td>
                  <td align="center">90.7</td>
                  <td align="center">66.7</td>
                </tr>
                <tr>
                  <td align="center">+ CBAM</td>
                  <td align="center">80.9</td>
                  <td align="center">89.3</td>
                  <td align="center">91.3</td>
                  <td align="center">87.2</td>
                  <td align="center">64.4</td>
                </tr>
                <tr>
                  <td align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">
                              <bold>+ Fusion Module</bold>
                            </td>
                          </tr>
                          <tr>
                            <td align="center">
                              <bold>(Algorithm 1)</bold>
                            </td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td align="center">
                    <bold>89.3</bold>
                  </td>
                  <td align="center">
                    <bold>94.1</bold>
                  </td>
                  <td align="center">
                    <bold>95.3</bold>
                  </td>
                  <td align="center">
                    <bold>92.9</bold>
                  </td>
                  <td align="center">
                    <bold>68.2</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" colspan="7" align="left">In the same group of experiments, the best-performing data is highlighted in bold.</td>
                </tr>
              </table>
            </table-wrap>
          </p>
        </sec>
        <sec id="S4.SS3.SSS2">
          <label>4.3.2</label>
          <title>Experimental Results Utilizing the Landsat 8-Based Sea Ice Dataset</title>
          <p id="S4.SS3.SSS2.p1">As shown in Table <xref rid="T6" ref-type="table">6</xref>, we conduct experiments on the landsat 8-based sea ice dataset with mainstream object detection algorithms and perform a comparative analysis of our improved YOLOv8. The selected object detection algorithms include: the two-stage object detection algorithm Faster R-CNN, the Transformer-based object detection algorithm RT-DETR, the one-stage object detection algorithms from the YOLO series, and the improved YOLO series algorithms. Compared to YOLOv8, our improved YOLOv8 achieves a 4.5% increase in Precision, a 1.1% improvement in mAP50, a 4.2% enhancement in mAP50-95, and a 1.5% boost in F1 score, while simultaneously reducing the training time by 51.0%.</p>
          <p id="S4.SS3.SSS2.p2">In addition, compared to other improved YOLO series algorithms, our enhanced YOLOv8 also demonstrates outstanding detection accuracy and relatively faster convergence speed. As Figure <xref ref-type="fig" rid="F15">15</xref> illustrates, we present the detection effects of our improved YOLOv8 alongside the baseline model YOLOv8.</p>
          <p>
            <fig-group>
              <fig id="F14.a">
                <caption>
                  <p>(a) YOLOv8 incorrectly predicts categories of sea ice.</p>
                  <graphic xlink:href="fig14a.pdf"/>
                </caption>
              </fig>
              <fig id="F14.b">
                <caption>
                  <p>(b) Our improved YOLOv8 correctly predicts the results.</p>
                  <graphic xlink:href="fig14b.pdf"/>
                </caption>
              </fig>
              <fig id="F14.c">
                <caption>
                  <p>(c) YOLOv8 incorrectly predicts categories of sea ice.</p>
                  <graphic xlink:href="fig14c.pdf"/>
                </caption>
              </fig>
              <fig id="F14.d">
                <caption>
                  <p>(d) Our improved YOLOv8 correctly predicts the results.</p>
                  <graphic xlink:href="fig14d.pdf"/>
                </caption>
              </fig>
              <fig id="F14">
                <label>Figure 15.</label>
                <caption>
                  <p>Schematic diagram of experiment results.</p>
                </caption>
              </fig>
            </fig-group>
          </p>
        </sec>
      </sec>
      <sec id="S4.SS4">
        <label>4.4</label>
        <title>Model Analyses</title>
        <sec id="S4.SS4.SSS1">
          <label>4.4.1</label>
          <title>Ablation Study</title>
          <p id="S4.SS4.SSS1.p1">As shown in Table <xref rid="T7" ref-type="table">7</xref>, we exhibit the results of ablation experiments based on our improved YOLOv8. On the basis of YOLOv8, we replace the Concat module with a fusion module to obtain Algorithm 1, we substitute the C-IoU with Shape-IoU to obtain Algorithm 2, and we add an evidence fusion module to obtain Algorithm 3.</p>
          <p id="S4.SS4.SSS1.p2">The experimental data from the NWPU-RESISC45 dataset reveal that our improved network architecture, incorporating a fusion module, improves the mAP50 of YOLOv8 by 1.7%. In addition, we propose an evidence fusion module that improves the mAP50 of YOLOv8 by 3.7%. The experimental data from Our Sea Ice Dataset reveal that our improved network architecture, incorporating a fusion module, improves the mAP50 of YOLOv8 by 0.2%. In addition, we propose an evidence fusion module that improves the mAP50 of YOLOv8 by 0.6%.</p>
        </sec>
        <sec id="S4.SS4.SSS2">
          <label>4.4.2</label>
          <title>Analyses for An Attention-Based Fusion Module</title>
          <p id="S4.SS4.SSS2.p1">As depicted in Table <xref rid="T8" ref-type="table">8</xref>, we present the outcomes of the ablation experiments conducted on YOLOv8, utilizing various mainstream attention mechanisms.</p>
          <p id="S4.SS4.SSS2.p2">As shown in Table <xref rid="T8" ref-type="table">8</xref>, the attention module affects the detection accuracy improvement of YOLOv8, while the fusion module compensates for the negative impact of using only the attention mechanism and enhances the overall detection performance.</p>
          <p id="S4.SS4.SSS2.p3">To intuitively perceive the positive impact of the fusion module, we use heatmaps to visualize the feature extraction effects of YOLOv8 after introducing our fusion module, as shown in Figure <xref ref-type="fig" rid="F16">16</xref>. Among them, Figure <xref ref-type="fig" rid="F16">16</xref> (a) shows the experimental results of our improved YOLOv8 on the NWPU-RESISC45 dataset, while Figure <xref ref-type="fig" rid="F16">16</xref> (b) presents the experimental results of our improved YOLOv8 on our sea ice dataset.</p>
          <p>
            <fig id="F16">
              <label>Figure 16.</label>
              <caption>
                <p>Schematic diagram of the heatmap.</p>
              </caption>
              <graphic xlink:href="fig15heat.pdf"/>
            </fig>
          </p>
        </sec>
        <sec id="S4.SS4.SSS3">
          <label>4.4.3</label>
          <title>Analyses for Loss Function</title>
          <p id="S4.SS4.SSS3.p1">We substitute the C-IoU [<xref rid="ref050" ref-type="bibr">50</xref>] loss with a contemporary mainstream loss function for bounding box regression. As demonstrated in Table <xref rid="T9" ref-type="table">9</xref>, we present the results of ablation experiments conducted using YOLOv8.</p>
          <p>
            <table-wrap id="T9">
              <label>Table 9</label>
              <caption>
                <p>Ablation study with loss function for bounding box regression.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">
                    <bold>Dataset</bold>
                  </td>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">
                    <bold>Method</bold>
                  </td>
                  <td style="border-top: 1px solid black;" colspan="3" align="center">
                    <bold>AP50(%)</bold>
                  </td>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">
                    <bold>mAP50 (%)</bold>
                  </td>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">
                              <bold>mAP</bold>
                            </td>
                          </tr>
                          <tr>
                            <td align="center">
                              <bold>50-95 (%)</bold>
                            </td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td style="border-top: 1px solid black;" rowspan="2" align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">
                              <bold>Training</bold>
                            </td>
                          </tr>
                          <tr>
                            <td align="center">
                              <bold>time (h)</bold>
                            </td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>small</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>medium</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>big</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="8" align="left">NWPU-RESISC45 [<xref rid="ref029" ref-type="bibr">29</xref>]</td>
                  <td style="border-top: 1px solid black;" align="center">YOLOv8 (C-IoU)</td>
                  <td style="border-top: 1px solid black;" align="center">80.3</td>
                  <td style="border-top: 1px solid black;" align="center">80.8</td>
                  <td style="border-top: 1px solid black;" align="center">83.7</td>
                  <td style="border-top: 1px solid black;" align="center">81.6</td>
                  <td style="border-top: 1px solid black;" align="center">56.2</td>
                  <td style="border-top: 1px solid black;" align="center">1.113</td>
                </tr>
                <tr>
                  <td align="center">+ G-IoU</td>
                  <td align="center">72.5</td>
                  <td align="center">75.9</td>
                  <td align="center">77.8</td>
                  <td align="center">75.4</td>
                  <td align="center">53.1</td>
                  <td align="center">1.058</td>
                </tr>
                <tr>
                  <td align="center">+ D-IoU</td>
                  <td align="center">72</td>
                  <td align="center">72.9</td>
                  <td align="center">75.3</td>
                  <td align="center">73.4</td>
                  <td align="center">56.2</td>
                  <td align="center">1.710</td>
                </tr>
                <tr>
                  <td align="center">+ F-IoU</td>
                  <td align="center">74</td>
                  <td align="center">75.1</td>
                  <td align="center">81.1</td>
                  <td align="center">76.7</td>
                  <td align="center">56</td>
                  <td align="center">1.071</td>
                </tr>
                <tr>
                  <td align="center">+ S-IoU</td>
                  <td align="center">74.8</td>
                  <td align="center">76</td>
                  <td align="center">82.9</td>
                  <td align="center">77.9</td>
                  <td align="center">55.6</td>
                  <td align="center">0.958</td>
                </tr>
                <tr>
                  <td align="center">+ W-IoU</td>
                  <td align="center">78.5</td>
                  <td align="center">79.8</td>
                  <td align="center">82</td>
                  <td align="center">80.1</td>
                  <td align="center">56.4</td>
                  <td align="center">1.000</td>
                </tr>
                <tr>
                  <td align="center">+ Inner-IoU</td>
                  <td align="center">75.3</td>
                  <td align="center">78.4</td>
                  <td align="center">78.8</td>
                  <td align="center">77.5</td>
                  <td align="center">53.9</td>
                  <td align="center">1.839</td>
                </tr>
                <tr>
                  <td align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">
                              <bold>+ Shape-IoU</bold>
                            </td>
                          </tr>
                          <tr>
                            <td align="center">
                              <bold>(Algorithm 2)</bold>
                            </td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td align="center">
                    <bold>82.3</bold>
                  </td>
                  <td align="center">
                    <bold>81.4</bold>
                  </td>
                  <td align="center">
                    <bold>84.7</bold>
                  </td>
                  <td align="center">
                    <bold>82.8</bold>
                  </td>
                  <td align="center">
                    <bold>58.3</bold>
                  </td>
                  <td align="center">
                    <bold>0.848</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;border-bottom: 1px solid black;" rowspan="8" align="left">Our Sea Ice Dataset</td>
                  <td style="border-top: 1px solid black;" align="center">YOLOv8 (C-IoU)</td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>89.1</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">94.0</td>
                  <td style="border-top: 1px solid black;" align="center">95.0</td>
                  <td style="border-top: 1px solid black;" align="center">92.7</td>
                  <td style="border-top: 1px solid black;" align="center">67.6</td>
                  <td style="border-top: 1px solid black;" align="center">9.435</td>
                </tr>
                <tr>
                  <td align="center">+ G-IoU</td>
                  <td align="center">80.9</td>
                  <td align="center">84.7</td>
                  <td align="center">83.7</td>
                  <td align="center">83.1</td>
                  <td align="center">66.4</td>
                  <td align="center">10.810</td>
                </tr>
                <tr>
                  <td align="center">+ D-IoU</td>
                  <td align="center">84.3</td>
                  <td align="center">85.5</td>
                  <td align="center">86</td>
                  <td align="center">85.3</td>
                  <td align="center">67.4</td>
                  <td align="center">10.138</td>
                </tr>
                <tr>
                  <td align="center">+ F-IoU</td>
                  <td align="center">82.9</td>
                  <td align="center">84.4</td>
                  <td align="center">83</td>
                  <td align="center">83.4</td>
                  <td align="center">68</td>
                  <td align="center">9.234</td>
                </tr>
                <tr>
                  <td align="center">+ S-IoU</td>
                  <td align="center">82</td>
                  <td align="center">87</td>
                  <td align="center">86.6</td>
                  <td align="center">85.2</td>
                  <td align="center">68.9</td>
                  <td align="center">10.465</td>
                </tr>
                <tr>
                  <td align="center">+ W-IoU</td>
                  <td align="center">83.7</td>
                  <td align="center">93.5</td>
                  <td align="center">93</td>
                  <td align="center">90.1</td>
                  <td align="center">68.7</td>
                  <td align="center">9.911</td>
                </tr>
                <tr>
                  <td align="center">+ Inner-IoU</td>
                  <td align="center">81.8</td>
                  <td align="center">86.5</td>
                  <td align="center">84.8</td>
                  <td align="center">84.4</td>
                  <td align="center">69</td>
                  <td align="center">11.868</td>
                </tr>
                <tr>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">
                              <bold>+ Shape-IoU</bold>
                            </td>
                          </tr>
                          <tr>
                            <td align="center">
                              <bold>(Algorithm 2)</bold>
                            </td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">84</td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>96.2</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>99.5</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>93.2</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>70.1</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>8.325</bold>
                  </td>
                </tr>
              </table>
            </table-wrap>
          </p>
          <p>
            <table-wrap id="T10">
              <label>Table 10</label>
              <caption>
                <p>Ablation study with evidence fusion.</p>
              </caption>
              <table>
                <tr>
                  <td style="border-top: 1px solid black;" align="left">
                    <bold>Dataset</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>Method</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>
                      <table>
                        <tr>
                          <td align="center">Small-Scale</td>
                        </tr>
                        <tr>
                          <td align="center">Sea Ice</td>
                        </tr>
                      </table>
                    </bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>
                      <table>
                        <tr>
                          <td align="center">Medium-Scale</td>
                        </tr>
                        <tr>
                          <td align="center">Sea Ice</td>
                        </tr>
                      </table>
                    </bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>
                      <table>
                        <tr>
                          <td align="center">Big-Scale</td>
                        </tr>
                        <tr>
                          <td align="center">Sea Ice</td>
                        </tr>
                      </table>
                    </bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>mAP50 (%)</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>mAP50-95 (%)</bold>
                  </td>
                  <td style="border-top: 1px solid black;" align="center">
                    <bold>Training Time (h)</bold>
                  </td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;" rowspan="2" align="left">NWPU- RESISC45 [<xref rid="ref029" ref-type="bibr">29</xref>]</td>
                  <td style="border-top: 1px solid black;" align="center">YOLOv8 (C-IoU)</td>
                  <td style="border-top: 1px solid black;" align="center">80.3</td>
                  <td style="border-top: 1px solid black;" align="center">80.8</td>
                  <td style="border-top: 1px solid black;" align="center">83.7</td>
                  <td style="border-top: 1px solid black;" align="center">81.6</td>
                  <td style="border-top: 1px solid black;" align="center">56.2</td>
                  <td style="border-top: 1px solid black;" align="center">1.113</td>
                </tr>
                <tr>
                  <td align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">
                              <bold>+ Evidence Fusion</bold>
                            </td>
                          </tr>
                          <tr>
                            <td align="center">
                              <bold>(Algorithm 2)</bold>
                            </td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td align="center">
                    <bold>83.9</bold>
                  </td>
                  <td align="center">
                    <bold>84.5</bold>
                  </td>
                  <td align="center">
                    <bold>87.5</bold>
                  </td>
                  <td align="center">
                    <bold>85.3</bold>
                  </td>
                  <td align="center">
                    <bold>56.8</bold>
                  </td>
                  <td align="center">-</td>
                </tr>
                <tr>
                  <td style="border-top: 1px solid black;border-bottom: 1px solid black;" rowspan="2" align="left">Our Sea Ice Dataset</td>
                  <td style="border-top: 1px solid black;" align="center">YOLOv8 (C-IoU)</td>
                  <td style="border-top: 1px solid black;" align="center">89.1</td>
                  <td style="border-top: 1px solid black;" align="center">94.0</td>
                  <td style="border-top: 1px solid black;" align="center">95.0</td>
                  <td style="border-top: 1px solid black;" align="center">92.7</td>
                  <td style="border-top: 1px solid black;" align="center">67.6</td>
                  <td style="border-top: 1px solid black;" align="center">9.435</td>
                </tr>
                <tr>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <p>
                      <table-wrap>
                        <table>
                          <tr>
                            <td align="center">
                              <bold>+ Evidence Fusion</bold>
                            </td>
                          </tr>
                          <tr>
                            <td align="center">
                              <bold>(Algorithm 2)</bold>
                            </td>
                          </tr>
                        </table>
                      </table-wrap>
                    </p>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>89.7</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>94.5</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>95.7</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>93.3</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">
                    <bold>68.0</bold>
                  </td>
                  <td style="border-bottom: 1px solid black;" align="center">-</td>
                </tr>
              </table>
            </table-wrap>
          </p>
          <p>
            <fig id="F17">
              <label>Figure 17.</label>
              <caption>
                <p>Schematic diagram of the loss function curve.</p>
              </caption>
              <graphic xlink:href="fig16loss.pdf"/>
            </fig>
          </p>
          <p id="S4.SS4.SSS3.p2">As illustrated in Table <xref rid="T9" ref-type="table">9</xref>, when compared to other enhanced methods, YOLOv8 utilizing Shape-IoU [<xref rid="ref041" ref-type="bibr">41</xref>] not only demonstrates superior detection accuracy across all three categories of sea ice, but it also significantly reduces convergence time.</p>
          <p id="S4.SS4.SSS3.p3">Compared to YOLOv8, our Algorithm 2, equipped with Shape-IoU, achieves 1.2% mAP50 improvement, 2.1% mAP50-95 improvement along with a reduction in training time of 23.8%, as validated by the NWPU-RESISC45 dataset. Furthermore, this YOLOv8 variant equipped with Shape-IoU realizes a 0.5% increase in mAP50 and a 2.5% rise in mAP50-95 while concurrently reducing training time by 11.8%, as confirmed by our sea ice dataset.</p>
          <p id="S4.SS4.SSS3.p4">To intuitively perceive the positive impact of the Shape-IoU, we use Loss function curve to visualize the convergence process of our improved YOLOv8 and YOLOv8, as shown in Figure <xref ref-type="fig" rid="F17">17</xref>. Among them, Figure <xref ref-type="fig" rid="F17">17</xref> (a) shows the convergence process of our improved YOLOv8 and YOLOv8 on the NWPU-RESISC45 dataset, while Figure <xref ref-type="fig" rid="F17">17</xref> (b) presents the convergence process of our improved YOLOv8 and YOLOv8 on our sea ice dataset.</p>
        </sec>
        <sec id="S4.SS4.SSS4">
          <label>4.4.4</label>
          <title>Analyses for Evidence Fusion</title>
          <p id="S4.SS4.SSS4.p1">YOLOv8 utilizes a detection architecture which separates the tasks of classification and localization. This architecture disassembles the tensors, enabling independent predictions for both the bounding box and the category of each target.</p>
          <p id="S4.SS4.SSS4.p2">The bounding boxes and categories of the targets in sea ice dataset are closely related. Consequently, the design of the detection architecture may result in inconsistencies between the predicted bounding boxes and their corresponding sea ice categories. For instance, a bounding box that represents large-scale sea ice might be inaccurately associated with a predicted category of medium-scale sea ice.</p>
          <p id="S4.SS4.SSS4.p3">In summary, we simultaneously transform the bounding box and category information predicted by YOLOv8 into multiple pieces of evidence that characterize uncertainty. Subsequently, we utilize an enhanced DSmT fusion inference algorithm to predict the new category. As shown in Table <xref rid="T10" ref-type="table">10</xref>, we exhibit the results of ablation experiments based on YOLOv8.</p>
          <p id="S4.SS4.SSS4.p4">From Table <xref rid="T10" ref-type="table">10</xref>, it is intuitively clear that YOLOv8, when using evidence fusion, achieves better detection accuracy on all three types of sea ice.</p>
        </sec>
      </sec>
    </sec>
    <sec id="S5">
      <label>5.</label>
      <title>Conclusion</title>
      <p id="S5.p1">In this paper, we propose a YOLOv8-based sea ice detection algorithm designed to identify sea ice of various sizes in satellite imagery. Firstly, we incorporate an attention-based fusion module into the concatenation component of the YOLOv8 neck network. Secondly, we substitute the C-IoU loss function in YOLOv8 with the more recent Shape-IoU as the boundary regression loss for the detection head. Thridly, we convert the inference results of the YOLOv8's output into uncertain multiple evidences according to the size distribution of sea ice in the dataset. Subsequently, we fuse multiple pieces of evidences and infer new results based on the improved DSmT fusion inference algorithm. These bring our improved YOLOv8, an sea ice detection algorithm for detecting sea ice of multiple sizes in satellite imagery. The results show that our improved YOLOv8 achieves the state-of-the-art performance in two aspects: identifying sea ice and dividing sea ice size compared with the baseline model and other advanced detection algorithms.</p>
    </sec>
  </body>
  <back>
    <ack>
      <title>Acknowledgments</title>
      <p id="ack.p1">This work was supported in part by the National Natural Science Foundation of China under Grant 62072392 and Grant 62272405; in part by the Shandong Natural Science Foundation of China under Grant ZR2020QF010; in part by the Yantai City Science and Technology Innovation Development Program - Basic Research Category Projects under Grant 2024JCYJ038.</p>
    </ack>
    <sec id="sec0100" sec-type="COI-statement">
      <title>Conflict of interest</title>
      <p>Yiping Luo is an employee of Deep Space Exploration Laboratory, Hefei 230000, China.</p>
    </sec>
    <ref-list>
      <title>References</title>
      <ref id="ref001">
        <label>[1]</label>
        <mixed-citation> Samset, B. H., Zhou, C., Fuglestvedt, J. S., Lund, M. T., Marotzke, J., &amp; Zelinka, M. D. (2023). Steady global surface warming from 1973 to 2022 but increased warming rate after 1990. <italic>Communications Earth </italic>&amp;<italic> Environment</italic>, 4(1), 400. [<uri>https://doi.org/10.1038/s43247-023-01061-4</uri>] </mixed-citation>
      </ref>
      <ref id="ref002">
        <label>[2]</label>
        <mixed-citation> McKay, D. I. A., Staal, A., Abrams, J. F., Winkelmann, R., Sakschewski, B., Loriani, S., … &amp; Lenton, T. M. (2022). Exceeding 1.5°C global warming could trigger multiple climate tipping points. <italic>Science</italic>, 377(6611), eabn7950. [<uri>https://doi.org/10.1126/science.abn7950</uri>] </mixed-citation>
      </ref>
      <ref id="ref003">
        <label>[3]</label>
        <mixed-citation> Screen, J. A., Deser, C., Smith, D. M., Zhang, X., Blackport, R., Kushner, P. J., … &amp; Sun, L. (2018). Consistency and discrepancy in the atmospheric response to Arctic sea-ice loss across climate models. <italic>Nature Geoscience</italic>, 11(3), 155-163. [<uri>https://doi.org/10.1038/s41561-018-0059-y</uri>] </mixed-citation>
      </ref>
      <ref id="ref004">
        <label>[4]</label>
        <mixed-citation> Li, H., &amp; Fedorov, A. (2021). Persistent freshening of the Arctic Ocean and changes in the North Atlantic salinity caused by Arctic sea ice decline. <italic>Climate Dynamics</italic>, 57(11), 2995-3013. [<uri>https://doi.org/10.1007/s00382-021-05850-5</uri>] </mixed-citation>
      </ref>
      <ref id="ref005">
        <label>[5]</label>
        <mixed-citation> Cao, Y., Liang, S., Sun, L., Liu, J., Cheng, X., Wang, D., … &amp; Feng, K. (2022). Trans-Arctic shipping routes expanding faster than the model projections. <italic>Global Environmental Change</italic>, 73, 102488. [<uri>https://doi.org/10.1016/j.gloenvcha.2022.102488</uri>] </mixed-citation>
      </ref>
      <ref id="ref006">
        <label>[6]</label>
        <mixed-citation> Min, C., Zhou, X., Luo, H., Yang, Y., Wang, Y., Zhang, J., &amp; Yang, Q. (2023). Toward quantifying the increasing accessibility of the Arctic Northeast Passage in the past four decades. <italic>Advances in Atmospheric Sciences</italic>, 40(12), 2378-2390. [<uri>https://doi.org/10.1007/s00376-022-2040-3</uri>] </mixed-citation>
      </ref>
      <ref id="ref007">
        <label>[7]</label>
        <mixed-citation> Kapsar, K., Gunn, G., Brigham, L., &amp; Liu, J. (2023). Mapping vessel traffic patterns in the ice-covered waters of the Pacific Arctic. <italic>Climatic Change</italic>, 176(7), 94. [<uri>https://doi.org/10.1007/s10584-023-03568-3</uri>]] </mixed-citation>
      </ref>
      <ref id="ref008">
        <label>[8]</label>
        <mixed-citation> Rodriguez Alvarez, N., Holt, B., Jaruwatanadilok, S., Podest, E., &amp; Cavanaugh, K. (2019). An Arctic sea ice multi-step classification based on GNSS-R data from the TDS-1 mission. <italic>Remote Sensing of Environment</italic>, 230, 111201. [<uri>https://doi.org/10.1016/j.rse.2019.05.021</uri>]] </mixed-citation>
      </ref>
      <ref id="ref009">
        <label>[9]</label>
        <mixed-citation> Cai, Y., Wan, F., Hu, S., &amp; Lang, S. (2022). Accurate prediction of ice surface and bottom boundary based on multi-scale feature fusion network. <italic>Applied Intelligence</italic>, 52(14), 16370-16381. [<uri>https://doi.org/10.1007/s10489-022-03362-1</uri>] </mixed-citation>
      </ref>
      <ref id="ref010">
        <label>[10]</label>
        <mixed-citation> Qaraqe, M., Yang, Y. D., Varghese, E. B., Elzein, A., &amp; Basaran, E. (2024). Crowd behavior detection: Leveraging video swin transformer for crowd size and violence level analysis. <italic>Applied Intelligence</italic>, 54(21), 10709-10730. [<uri>https://doi.org/10.1007/s10489-024-05775-6</uri>] </mixed-citation>
      </ref>
      <ref id="ref011">
        <label>[11]</label>
        <mixed-citation> Li, X., Zhou, Y., Du, P., Lang, G., Xu, M., &amp; Wu, W. (2021). A deep learning system that generates quantitative CT reports for diagnosing pulmonary Tuberculosis. <italic>Applied Intelligence</italic>, 51(6), 4082-4093. [<uri>https://doi.org/10.1007/s10489-020-02051-1</uri>] </mixed-citation>
      </ref>
      <ref id="ref012">
        <label>[12]</label>
        <mixed-citation> Knausgård, K., Wiklund, A., Sørdalen, T., Halvorsen, K., Kleiven, A., Jiao, L., &amp; Goodwin, M. (2022). Temperate fish detection and classification: A deep learning based approach. <italic>Applied Intelligence</italic>, 52(6), 6988-7001. [<uri>https://doi.org/10.1007/s10489-020-02154-9</uri>] </mixed-citation>
      </ref>
      <ref id="ref013">
        <label>[13]</label>
        <mixed-citation> Girshick, R., Donahue, J., Darrell, T., &amp; Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. In <italic>2014 IEEE Conference on Computer Vision and Pattern Recognition</italic> (pp. 580-587). [<uri>https://doi.org/10.1109/CVPR.2014.81</uri>] </mixed-citation>
      </ref>
      <ref id="ref014">
        <label>[14]</label>
        <mixed-citation> Girshick, R. (2015). Fast R-CNN. In <italic>2015 IEEE International Conference on Computer Vision</italic> (pp. 1440-1448). [<uri>https://doi.org/10.1109/ICCV.2015.169</uri>] </mixed-citation>
      </ref>
      <ref id="ref015">
        <label>[15]</label>
        <mixed-citation> Ren, S., He, K., Girshick, R., &amp; Sun, J. (2017). Faster R-CNN: Towards real-time object detection with region proposal networks. <italic>IEEE Transactions on Pattern Analysis and Machine Intelligence</italic>, 39(6), 1137-1149. [<uri>https://doi.org/10.1109/TPAMI.2016.2577031</uri>] </mixed-citation>
      </ref>
      <ref id="ref016">
        <label>[16]</label>
        <mixed-citation> He, K., Gkioxari, G., Dollár, P., &amp; Girshick, R. (2017). Mask R-CNN. In <italic>2017 IEEE International Conference on Computer Vision</italic> (pp. 2961-2969). [<uri>https://doi.org/10.1109/ICCV.2017.322</uri>] </mixed-citation>
      </ref>
      <ref id="ref017">
        <label>[17]</label>
        <mixed-citation> Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C. Y., &amp; Berg, A. C. (2016). Ssd: Single shot multibox detector. In <italic>Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14 </italic>(pp. 21-37). Springer International Publishing. [<uri>https://doi.org/10.1007/978-3-319-46448-0_2</uri>] </mixed-citation>
      </ref>
      <ref id="ref018">
        <label>[18]</label>
        <mixed-citation> Redmon, J., Divvala, S., Girshick, R., &amp; Farhadi, A. (2016). You only look once: Unified, real-time object detection. In <italic>2016 IEEE Conference on Computer Vision and Pattern Recognition</italic> (pp. 779-788). [<uri>https://doi.org/10.1109/CVPR.2016.91</uri>] </mixed-citation>
      </ref>
      <ref id="ref019">
        <label>[19]</label>
        <mixed-citation> Redmon, J., &amp; Farhadi, A. (2017). YOLO9000: Better, faster, stronger. In <italic>2017 IEEE Conference on Computer Vision and Pattern Recognition</italic> (pp. 7263-7271). [<uri>https://doi.org/10.1109/CVPR.2017.690</uri>] </mixed-citation>
      </ref>
      <ref id="ref020">
        <label>[20]</label>
        <mixed-citation> Redmon, J., &amp; Farhadi, A. (2018). YOLOv3: An incremental improvement. <italic>arXiv preprint arXiv:1804.02767</italic>. </mixed-citation>
      </ref>
      <ref id="ref021">
        <label>[21]</label>
        <mixed-citation> Bochkovskiy, A., Wang, C. Y., &amp; Liao, H. Y. M. (2020). YOLOv4: Optimal speed and accuracy of object detection. <italic>arXiv preprint arXiv:2004.10934</italic>. </mixed-citation>
      </ref>
      <ref id="ref022">
        <label>[22]</label>
        <mixed-citation> Jocher, G., Stoken, A., Borovec, J., Changyu, L., Hogan, A., Diaconu, L., … &amp; Dave, P. (2020). ultralytics/yolov5: v3. 0. Zenodo. Retrieved from <ext-link xlink:href="https://ui.adsabs.harvard.edu/link_gateway/2020zndo...3983579J/doi:10.5281/zenodo.3983579">https://ui.adsabs.harvard.edu/link_gateway/2020zndo...3983579J/doi:10.5281/zenodo.3983579</ext-link> </mixed-citation>
      </ref>
      <ref id="ref023">
        <label>[23]</label>
        <mixed-citation> Li, C., Li, L., Geng, Y., Jiang, H., Cheng, M., Zhang, B., … &amp; Chu, X. (2023). YOLOv6 v3.0: A full-scale reloading. <italic>arXiv preprint arXiv:2301.05586</italic>. </mixed-citation>
      </ref>
      <ref id="ref024">
        <label>[24]</label>
        <mixed-citation> Wang, C. Y., Bochkovskiy, A., &amp; Liao, H. Y. M. (2023). YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. In <italic>2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition</italic> (pp. 7464-7475). [<uri>https://doi.org/10.1109/CVPR52729.2023.00721</uri>] </mixed-citation>
      </ref>
      <ref id="ref025">
        <label>[25]</label>
        <mixed-citation> Jocher, G., Chaurasia, A., &amp; Qiu, J. (2023). Ultralytics YOLOv8. GitHub repository. Retrieved from <ext-link xlink:href="https://github.com/ultralytics/ultralytics">https://github.com/ultralytics/ultralytics</ext-link> </mixed-citation>
      </ref>
      <ref id="ref026">
        <label>[26]</label>
        <mixed-citation> Wang, C. Y., Yeh, I. H., &amp; Liao, H. (2024). YOLOv9: Learning what you want to learn using programmable gradient information. <italic>arXiv preprint arXiv:2402.13616</italic>. </mixed-citation>
      </ref>
      <ref id="ref027">
        <label>[27]</label>
        <mixed-citation> Wang, A., Chen, H., Liu, L., Chen, K., Lin, Z., Han, J., &amp; Ding, G. (2024). YOLOv10: Real-time end-to-end object detection. <italic>arXiv preprint arXiv:2405.14458</italic>. </mixed-citation>
      </ref>
      <ref id="ref028">
        <label>[28]</label>
        <mixed-citation> Li, W., Hsu, C. Y., &amp; Tedesco, M. (2024). Advancing Arctic sea ice remote sensing with AI and deep learning: Opportunities and challenges. <italic>Remote Sensing</italic>, 16(20), 3764. [<uri>https://doi.org/10.3390/rs16203764</uri>] </mixed-citation>
      </ref>
      <ref id="ref029">
        <label>[29]</label>
        <mixed-citation> Cheng, G., Han, J., &amp; Lu, X. (2017). Remote sensing image scene classification: Benchmark and state of the art. <italic>Proceedings of the IEEE</italic>, 105(10), 1865-1883. [<uri>https://doi.org/10.1109/JPROC.2017.2675998</uri>] </mixed-citation>
      </ref>
      <ref id="ref030">
        <label>[30]</label>
        <mixed-citation> Rogers, M., Fox, M., Fleming, A., Zeeland, L., Wilkinson, J., &amp; Hosking, S. (2024). Sea ice detection using concurrent multispectral and synthetic aperture radar imagery. <italic>Remote Sensing of Environment</italic>, 305, 114073. [<uri>https://doi.org/10.1016/j.rse.2024.114073</uri>] </mixed-citation>
      </ref>
      <ref id="ref031">
        <label>[31]</label>
        <mixed-citation> Sandven, S., Spreen, G., Heygster, G., Girard-Ardhuin, F., Farrell, S., Dierking, W., &amp; Allard, R. (2023). Sea ice remote sensing—Recent developments in methods and climate data sets. <italic>Surveys in Geophysics</italic>, 44(5), 1653-1689. [<uri>https://doi.org/10.1007/s10712-023-09781-0</uri>] </mixed-citation>
      </ref>
      <ref id="ref032">
        <label>[32]</label>
        <mixed-citation> Hu, Y., Hua, X., Yan, Q., Liu, W., Jiang, Z., &amp; Wickert, J. (2024). Sea ice detection from GNSS-R data based on local linear embedding. <italic>Remote Sensing</italic>, 16(14), 2621. [<uri>https://doi.org/10.3390/rs16142621</uri>] </mixed-citation>
      </ref>
      <ref id="ref033">
        <label>[33]</label>
        <mixed-citation> Liu, L., Dong, X., Lin, W., &amp; Lang, S. (2023). Polar sea ice detection using a rotating fan beam scatterometer. <italic>Remote Sensing</italic>, 15(20), 5063. [<uri>https://doi.org/10.3390/rs15205063</uri>] </mixed-citation>
      </ref>
      <ref id="ref034">
        <label>[34]</label>
        <mixed-citation> Jafari, Z., Bobby, P., Karami, E., &amp; Taylor, R. (2025). Machine learning-based detection of icebergs in sea ice and open water using SAR imagery. <italic>Remote Sensing</italic>, 17(4), 702. [<uri>https://doi.org/10.3390/rs17040702</uri>] </mixed-citation>
      </ref>
      <ref id="ref035">
        <label>[35]</label>
        <mixed-citation> Xiong, Y., Wang, D., Fu, D., &amp; Huang, H. (2023). Ice identification with error-accumulation enhanced neural dynamics in optical remote sensing images. <italic>Remote Sensing</italic>, 15(23), 5555. [<uri>https://doi.org/10.3390/rs15235555</uri>] </mixed-citation>
      </ref>
      <ref id="ref036">
        <label>[36]</label>
        <mixed-citation> Chai, Y., Ren, J., Hwang, B., Wang, J., Fan, D., Yan, Y., &amp; Zhu, S. (2021). Texture-sensitive superpixeling and adaptive thresholding for effective segmentation of sea ice floes in high-resolution optical images. <italic>IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing</italic>, 14, 577-586. [<uri>https://doi.org/10.1109/JSTARS.2020.3040614</uri>] </mixed-citation>
      </ref>
      <ref id="ref037">
        <label>[37]</label>
        <mixed-citation> Qiu, Y., Li, X. M., &amp; Guo, H. (2023). Spaceborne thermal infrared observations of Arctic sea ice leads at 30 m resolution. <italic>The Cryosphere</italic>, 17(7), 2829-2849. [<uri>https://doi.org/10.5194/tc-17-2829-2023</uri>] </mixed-citation>
      </ref>
      <ref id="ref038">
        <label>[38]</label>
        <mixed-citation> Liang, S., Zeng, J. Y., Li, Z., Chen, K. S., &amp; Zhang, P. (2020). Assessment of four passive microwave sea ice concentrations by using automatic MODIS sea ice classification. In <italic>IGARSS 2020-2020 IEEE International Geoscience and Remote Sensing Symposium</italic> (pp. 3039-3042). [<uri>https://doi.org/10.1109/IGARSS39084.2020.9324590</uri>] </mixed-citation>
      </ref>
      <ref id="ref039">
        <label>[39]</label>
        <mixed-citation> Ding, S., Zeng, D., Zhou, L., Han, S., Li, F., &amp; Wang, Q. (2023). Multi-scale polar object detection based on computer vision. <italic>Water</italic>, 15(19), 3431. [<uri>https://doi.org/10.3390/w15193431</uri>] </mixed-citation>
      </ref>
      <ref id="ref040">
        <label>[40]</label>
        <mixed-citation> Hu, J., Shen, L., &amp; Sun, G. (2018). Squeeze-and-excitation networks. In <italic>2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition</italic> (pp. 7132-7141). [<uri>https://doi.org/10.1109/CVPR.2018.00745</uri>] </mixed-citation>
      </ref>
      <ref id="ref041">
        <label>[41]</label>
        <mixed-citation> Zhang, H., &amp; Zhang, S. (2023). Shape-IoU: More accurate metric considering bounding box shape and scale. <italic>arXiv preprint arXiv:2312.17663</italic>. </mixed-citation>
      </ref>
      <ref id="ref042">
        <label>[42]</label>
        <mixed-citation> Guo, Q., Pan, X. &amp; Tang, T. (2023). DSmT-DS Multi-Source Uncertainty Reasoning Methodology. <italic>Multi-source Uncertain Information Reasoning Technology</italic>, (pp. 59-60). </mixed-citation>
      </ref>
      <ref id="ref043">
        <label>[43]</label>
        <mixed-citation> Bahdanau, D., Cho, K., &amp; Bengio, Y. (2014). Neural machine translation by jointly learning to align and translate. <italic>arXiv preprint arXiv:1409.0473</italic>. </mixed-citation>
      </ref>
      <ref id="ref044">
        <label>[44]</label>
        <mixed-citation> Jaderberg, M., Simonyan, K., Zisserman, A., &amp; Kavukcuoglu, K. (2015). Spatial transformer networks. <italic>arXiv preprint arXiv:1506.02025</italic>. </mixed-citation>
      </ref>
      <ref id="ref045">
        <label>[45]</label>
        <mixed-citation> Woo, S., Park, J., Lee, J. Y., &amp; Kweon, I. S. (2018). CBAM: Convolutional block attention module. In <italic>Proceedings of the European Conference on Computer Vision</italic> (pp. 3-19). [<uri>https://doi.org/10.1007/978-3-030-01234-2_1</uri>] </mixed-citation>
      </ref>
      <ref id="ref046">
        <label>[46]</label>
        <mixed-citation> Wang, C. Y., Liao, H. Y. M., Wu, Y. H., Chen, P. Y., Hsieh, J. W., &amp; Yeh, I. H. (2020). CSPNet: A new backbone that can enhance learning capability of CNN. In <italic>2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops</italic> (pp. 1571-1580). [<uri>https://doi.org/10.1109/CVPRW50498.2020.00203</uri>] </mixed-citation>
      </ref>
      <ref id="ref047">
        <label>[47]</label>
        <mixed-citation> Zhang, X., Zeng, H., Guo, S., &amp; Zhang, L. (2022). Efficient long-range attention network for image super-resolution. In <italic>European Conference on Computer Vision</italic> (pp. 649-667). [<uri>https://doi.org/10.1007/978-3-031-19790-1_39</uri>] </mixed-citation>
      </ref>
      <ref id="ref048">
        <label>[48]</label>
        <mixed-citation> Yu, J., Jiang, Y., Wang, Z., Cao, Z., &amp; Huang, T. S. (2016). UnitBox: An advanced object detection network. In <italic>Proceedings of the 24th ACM International Conference on Multimedia</italic> (pp. 516-520). [<uri>https://doi.org/10.1145/2964284.2967274</uri>] </mixed-citation>
      </ref>
      <ref id="ref049">
        <label>[49]</label>
        <mixed-citation> Rezatofighi, H., Tsoi, N., Gwak, J. Y., Sadeghian, A., Reid, I., &amp; Savarese, S. (2019). Generalized intersection over union: A metric and a loss for bounding box regression. In <italic>2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition</italic> (pp. 658-666). [<uri>https://doi.org/10.1109/CVPR.2019.00075</uri>] </mixed-citation>
      </ref>
      <ref id="ref050">
        <label>[50]</label>
        <mixed-citation> Zheng, Z., Wang, P., Liu, W., Li, J., Ye, R., &amp; Ren, D. (2020). Distance-IoU loss: Faster and better learning for bounding box regression. <italic>Proceedings of the AAAI Conference on Artificial Intelligence</italic>, 34(07), 12993-13000. [<uri>https://doi.org/10.1609/aaai.v34i07.6999</uri>] </mixed-citation>
      </ref>
      <ref id="ref051">
        <label>[51]</label>
        <mixed-citation> Zhang, Y. F., Ren, W., Zhang, Z., Jia, Z., Wang, L., &amp; Tan, T. (2022). Focal and efficient IOU loss for accurate bounding box regression. <italic>Neurocomputing</italic>, 506, 146-157. [<uri>https://doi.org/10.1016/j.neucom.2022.07.042</uri>] </mixed-citation>
      </ref>
      <ref id="ref052">
        <label>[52]</label>
        <mixed-citation> Gevorgyan, Z. (2022). SIoU loss: More powerful learning for bounding box regression. <italic>arXiv preprint arXiv:2205.12740</italic>. </mixed-citation>
      </ref>
      <ref id="ref053">
        <label>[53]</label>
        <mixed-citation> Tong, Z., Chen, Y., Xu, Z., &amp; Yu, R. (2023). Wise-IoU: Bounding box regression loss with dynamic focusing mechanism. <italic>arXiv preprint arXiv:2301.10051</italic>. </mixed-citation>
      </ref>
      <ref id="ref054">
        <label>[54]</label>
        <mixed-citation> Zhang, H., Xu, C., &amp; Zhang, S. (2023). Inner-IoU: More effective intersection over union loss with auxiliary bounding box. <italic>arXiv preprint arXiv:2311.02877</italic>. </mixed-citation>
      </ref>
      <ref id="ref055">
        <label>[55]</label>
        <mixed-citation> Tzutalin. (2021). LabelImg. <italic>PyPI</italic>. Retrieved from <ext-link xlink:href="https://pypi.org/project/labelImg/">https://pypi.org/project/labelImg/</ext-link> </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>
