Fast Detection of Tomato Peduncle Using Point Cloud with a Harvesting Robot

This paper proposes a fast method for detecting tomato peduncles by a harvesting robot. The main objective of this study is to develop automated harvesting with a robot. The harvesting robot is equipped with an RGB-D camera to detect peduncles, and an end effector to harvest tomatoes. It is necessary for robots to detect where to cut a plant for harvesting. The proposed method detects peduncles using a point cloud created by the RGB-D camera. Pre-processing is performed with voxelization in two resolutions to reduce the computational time needed to calculate the positional relationship between voxels. Finally, an energy function is defined based on three conditions of a peduncle, and this function is minimized to identify the cutting point on each peduncle. To experimentally demonstrate the effectiveness of our approach, a robot was used to identify the peduncles of target tomato plants and harvest the tomatoes at a real farm. Using the proposed method, the harvesting robot achieved peduncle detection of the tomatoes, and harvested tomatoes successfully by cutting the peduncles.