A cherry tomato classification-picking Robot based on the K-means algorithm

Abstract

In order to improve the automation level of cherry tomato harvesting, this paper proposes an cherry tomato picking robot. The robot includes four parts: A vision camera, a mechanical arm, a picking claw and a tracked AGV. The vision camera learns and classifies different varieties of cherry tomatoes based on the K-means algorithm. The color patch-based visual tracking algorithm accurately locates the ripe tomatoes, and the robotic arm drives the claws to pick them accurately. The tracked AGV adopts the original difference height intelligent tracking method, which has low cost and high adaptability to the working environment. It is verified by experiment that the cherry tomato classification-picking robot can effectively identify and locate different types of cherry tomatoes. The single-fruit tomato picking operation takes about 20s, the success rate is over 80%.