Cherry Tomato Harvesting Robot with 3-D Sensor on Its End Effector (Part 2)

Image Recognition and Harvest Experiment

Abstract

A cherry tomato harvesting robot equipped with a 3-D vision sensor on the end effector was made for trial purpose. Image recognition and harvest experiments were carried out in a green house. The 3-D vision sensor scanned the front field of the robot. Red fruits and obstacles in the field were recognized by processing the image. These fruits were picked by the end effector avoiding a collision with obstacles. Furthermore, this image was processed to find the space among the stalks and leaves for scanning the side field, and to find the dead angle in its image. If both of them were found, the 3-D vision sensor was controlled to scan the side field. The red fruits recognized by processing the side image were also picked by the end effector. Result of the experiment showed that most of the experimented red fruits were recognized and harvested, although some problems must be solved to decrease the dropping of the fruits and to harvest the fruit with its calyx.