Intelligent Tomato Picking Robot System Based on Multimodal Depth Feature Analysis Method

Abstract

In order to achieve monomer picking and string picking of tomato fruit. The tomato image was obtained by binocular camera, and the preliminary search of tomato fruit was carried out. Then the second close-up camera was used to obtain the accurate image of tomato fruit. The improved Deeplab model was used to complete the recognition and localization of tomato fruit, and the fruit was detected by Harris algorithm. The corner point of the fruit stem of the first fruit bifurcation point and the stem of the plant is determined as the position of the cut point of the tomato fruit stem. The robotic arm, the end effector and the mechanical knife are controlled to complete the picking of the individual tomato fruit and the string of tomato fruits, providing a solution for the manual picking of the replacement tomato.