DYNAMIC POTATO IDENTIFICATION AND CLEANING METHOD BASED ON RGB-D

To solve the problems of a large number clods remaining in potatoes after mechanized harvesting northern heavy clay soil planting areas China and requiring much labor to separate from potatoes, which leads workload, inefficiency low cleaning rate, an RGB-D-based Mask R-CNN dynamic potato identification model is established by using acquired RBG-D image data untreated harvesting, method presented this paper. This makes it possible automatically clod impurities potatoes. The experimental results showed that prediction accuracy more than 97%. With increase conveyance speed, actual precision show downward trend. Comprehensively considering efficiency accuracy, when speed set 0.4 m·s−1, reaches as high 96.35%. research provides theoretical reference for further study intelligent systems.