Study on recognition and non-destructive picking end-effector of kiwifruit

Fruit information perception and nondestructive picking method are the key technologies for harvesting robot. In this paper, the kiwifruit automatic identification method and nondestructive picking end-effector were studied. Firstly, a method of obtaining image from the bottom of fruit was proposed based on the clustering characteristics of fruit and the method of elliptic Hough transform was used to make the target fruit be recognized separately in this procession to determine the picking order of fruit. Secondly, an end-effector was designed and manufactured, which could approach the fruit from the bottom, enveloping and grabbing the fruit from two sides, and rotated the fruit until separating the fruit and the stem. Finally, a comprehensive experiment was made combined with kiwifruit automatic identification method and nondestructive picking end-effector. Also, a harvesting adaptability research was finished by identification method. The result shows that, the recognition rate was 95.74% and the picking success rate was 88.89%.