Tomato Harvesting Robot System Based on Binocular Vision

Abstract

A harvesting robot system was created in this paper based on binocular vision, in order to reduce the amount of human labor. With the use of the deep learning, tomatoes could be easily detected. Based on the proposed stereo matching algorithm, the localization of each target can be computed accurately. A tiny finger shaped gripper was also designed to separate the tomatoes from the stem. Finally, the performance of the robot system was evaluated. The result showed that the successful harvesting rate was 82%. And the whole cycle of a harvesting cost 13.2 seconds.