2P1-C06 Vision-Based Behavior Strategy for Tomato Harvesting Robot : 3D Reconstruction from Multi-images of Tomato bunch and Fruits' Pose Estimation

Abstract

This paper presents a visual information aided behavior strategy for tomato harvesting robot. Particularly, we have proposed an image analysis method for estimating the pose of the fruits as it is the most essential information for picking tomato. Tomato harvesting robot includes a six-axis serial manipulator with a monocular hand-eye and an end-effector, and autonomous mobile cart along with a digital computer. Multi-images of tomato bunch captured from different views by hand-eye are used to reconstruct a 3D point cloud of the bunch which provides the basic information for pose estimation. This information is used in the proposed algorithm to estimate the pose of tomatoes with different shapes and sizes.