TY  - JOUR
AU  - Zhang, Zhien
AU  - Chen, Weiwei
AU  - Huang, Huichan
AU  - Dai, Shenyuan
PY  - 2026
DA  - 2026/06/29
TI  - Design and Experiments of a Machine Vision-Based Directional Arrangement Device for Packaged Vegetables
JO  - Digital Intelligence in Agriculture
T2  - Digital Intelligence in Agriculture
JF  - Digital Intelligence in Agriculture
VL  - 2
IS  - 2
SP  - 103
EP  - 113
DO  - 10.62762/DIA.2026.219926
UR  - https://www.icck.org/article/abs/DIA.2026.219926
KW  - machine vision
KW  - plant factory
KW  - mechanism design
KW  - head-tail orientation
KW  - prototype testbed
AB  - Plant factories, as agricultural production systems characterized by high yield, efficient resource utilization, and advanced mechanization, have attracted increasing global attention. Packaging finished vegetables is a critical pre-shipment operation in plant-factory production, and its automation and intelligent control remain urgent research needs. In vegetable-packaging line, an orientation-adjustment mechanism is required to correct the posture of packaged vegetables so that the boxing mechanism can place them neatly in turnover boxes. The operation of this mechanism depends on accurate identification of vegetable orientation. In this study, packaged vegetables produced in a plant factory were selected as the research object. Their postures on a conveyor belt were simulated, and images were acquired for orientation recognition. Multiple image features and classifiers were compared, and suitable texture features and classifiers were selected to identify the orientation of packaged vegetables. A translating-oscillating compound cam mechanism was designed according to the motion requirements for flipping packaged vegetables. The design process included determination of the follower motion law, generation of the cam profile curve, establishment of the cam model, and optimization of the structural parameters through a human-computer interaction framework. A prototype was then fabricated and assembled, and tests were conducted to verify the accuracy of the visual recognition system and the feasibility of the steering mechanism. The results showed that the GA-SVM classifier achieved a maximum accuracy of 97.90% with the fused HOG and GLCM features. The average accuracy of the recognition model was 95.27%, and the success rate of the steering mechanism was 94.67%. Both the recognition and steering performances were satisfactory. The proposed orientation-identification method and turning mechanism for packaged vegetables in plant factories were successfully implemented, providing a technical basis for unmanned plant-factory production lines.
SN  - 3069-3187
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Zhang2026Design,
  author = {Zhien Zhang and Weiwei Chen and Huichan Huang and Shenyuan Dai},
  title = {Design and Experiments of a Machine Vision-Based Directional Arrangement Device for Packaged Vegetables},
  journal = {Digital Intelligence in Agriculture},
  year = {2026},
  volume = {2},
  number = {2},
  pages = {103-113},
  doi = {10.62762/DIA.2026.219926},
  url = {https://www.icck.org/article/abs/DIA.2026.219926},
  abstract = {Plant factories, as agricultural production systems characterized by high yield, efficient resource utilization, and advanced mechanization, have attracted increasing global attention. Packaging finished vegetables is a critical pre-shipment operation in plant-factory production, and its automation and intelligent control remain urgent research needs. In vegetable-packaging line, an orientation-adjustment mechanism is required to correct the posture of packaged vegetables so that the boxing mechanism can place them neatly in turnover boxes. The operation of this mechanism depends on accurate identification of vegetable orientation. In this study, packaged vegetables produced in a plant factory were selected as the research object. Their postures on a conveyor belt were simulated, and images were acquired for orientation recognition. Multiple image features and classifiers were compared, and suitable texture features and classifiers were selected to identify the orientation of packaged vegetables. A translating-oscillating compound cam mechanism was designed according to the motion requirements for flipping packaged vegetables. The design process included determination of the follower motion law, generation of the cam profile curve, establishment of the cam model, and optimization of the structural parameters through a human-computer interaction framework. A prototype was then fabricated and assembled, and tests were conducted to verify the accuracy of the visual recognition system and the feasibility of the steering mechanism. The results showed that the GA-SVM classifier achieved a maximum accuracy of 97.90\% with the fused HOG and GLCM features. The average accuracy of the recognition model was 95.27\%, and the success rate of the steering mechanism was 94.67\%. Both the recognition and steering performances were satisfactory. The proposed orientation-identification method and turning mechanism for packaged vegetables in plant factories were successfully implemented, providing a technical basis for unmanned plant-factory production lines.},
  keywords = {machine vision, plant factory, mechanism design, head-tail orientation, prototype testbed},
  issn = {3069-3187},
  publisher = {Institute of Central Computation and Knowledge}
}