关键词: 机器人, 角点检测, 曲率, 图像处理, 阈值分割

Abstract: The task of apple picking robot become more complicated for swing and overlapping of
apple. Thus the paper proposes a new apple picking method based on visual identification and
location technology. It can help robot to find out target more efficiently by visual inspection
technology in the Orchard. Firstly, we applied image filtering, histogram equalization and image
segmentation to make the background and fruit’s contour separated. Secondly, the proposed algorithm
applied to calculate the curvature of edge contour to find the peak point of curvature. Finally, this
study converted the pixel coordinate of peak point calibrated, and turning pixel coordinate into world
coordinate system to make it the target point for robot which robot can grab. Robot could adjust
posture to trace fruit and identify it with position information of target point, and then find it, grab it,
and pick it accurately. The experiment showed the robot grab the fruits of success rate as high as
ninety percent above by applying proposed method. The proposed study also provide some valuable
reference for identify and locating of spherical or round shaped fruits.

Key words: robot, corner detection, curvature, image processing, threshold segmentation