Reconstruction Method of Overlapped Citrus Fruits in Natural Scenes Based on Convex Hull

A reconstruction method based on convex hull was developed to reconstruct the overlapped citrus in natural scenes in order to improve the success rate of citrus harvesting robot. Due to the circular shape of citrus fruit, two reconstruction parameters were determined: the center point and the radius. Firstly, the K-means clustering method and threshold segmentation method were applied to detect the citrus targets in HSV color space. After that, the center points of each citrus fruit targets were extracted respectively by combing with the convex hull and the minimum external affine transformation rectangle. Finally, the reconstruction radiuses of the overlapped citrus targets were determined by searching the gray scale gradient extremum along the connecting line of each center point of citrus and the occlusion order is given. Experiments verified the validity of the method proposed. It provides a basis for the design of end-effector and the path planning of the harvesting robot.