An Automated Equipment Design for Visual Inspection and Rapid Sorting of Agricultural Products

In response to the problems such as low manual sorting efficiency, high loss rate and inconsistent standards in the Zizania latifolia industry in Chang shun County, Qian nan Prefecture, Guizhou Province, an automated equipment based on visual inspection and mechanical sorting has been designed. Through intelligent technology, the efficiency and accuracy of Zizania latifolia sorting are improved, which helps rural revitalization. Firstly, this study analyzes the physical characteristics and sorting requirements of Zizania latifolia, and proposes an automated sorting system composed of a visual recognition module, a mechanical sorting module and a control system. In terms of the visual recognition algorithm, deep learning models such as MobileNetV2, VGG16 and EfficientNet are compared. Finally, MobileNetV2 is selected as the core algorithm. Through transfer learning and data augmentation techniques, high-precision classification of the quality of Zizania latifolia is achieved, with an accuracy rate of over 97.2%. In terms of mechanical design, a SCARA-class robotic arm and flexible grippers are adopted to ensure the flexibility and accuracy of the sorting process. The sorting efficiency is increased to 2.99 seconds per piece, and the loss rate is reduced to below 5%. The results show that this equipment is superior to manual sorting in terms of both sorting efficiency and grading accuracy, and it has the ability to operate around the clock, which can meet the needs of large-scale production. This study provides a feasible technical solution for the automated sorting of agricultural products, and it is of great significance for promoting the intelligent development of agriculture.