Development of a Tomato Harvesting Robot with a Vision System Using Specular Reflection (Part 1)

——Selection of Hand-Approach Direction by Image Processing——

Abstract

With the aim of achieving automation of tomato harvesting, the authors prototyped a detecting system that selects the approach direction for the harvesting hand and clarify the accuracy. In operation, the processing program first extracts an entire tomato cluster, which consists of red colored fruit and green colored fruit. Next, the pixels of 9 meshes—around and to the forward direction—of the nearest red colored tomato to the hand are calculated. The direction from which the hand can easily grasp the tomato is selected automatically based on the pixel information. Image-processing experiments have shown that in 89-96% of cases, the system succeeded in determining an approach direction for clusters bearing 2-4 tomatoes.