房状小果実のロボット収穫のための位置検出アルゴリズム

人間の視覚と比較して

抄録

Several methods to detect fruit position for robotic harvesting have been proposed. They are classified into active range finder and passive range finder, however both methods are based on principle of triangulation. In this study, a robotic system for harvesting small fruits in truss using passive range finder was investigated. The main tasks of this study was to design an efficient and effective machine vision based guidance system when an harvesting end-effector which could blow fruit into a fruit transporting tube was used. The challenge of cherry tomato fruit detection was to determine all fruit positions precisely, because of the corresponding problem of matching multiple objects in the two images of a stereo image set. In fact, the 3-D position of fruit cluster was determined using a binocular stereo vision technique. The robot harvested one fruit at a time and the position of the next target fruit was updated based on a newly acquired image and the latest manipulator position. The experimental results showed that this visual feedback control based harvesting method was effective. In addition, it was observed that the movements of fruits due to the harvest action or air movement could be compensated for to assure successful harvest and that the algorithm of this method was very similar with that of human being's eyesight.