An Algorithm for Calculating Apple Picking Direction Based on 3D Vision

Research into autonomous (robotic) apple picking has not yet resolved the problem of finding optimal orientation. Robotic efficiency, in terms all available apples without loss or damage, remains low. This paper proposes a method determining orientation relative to target fruit and adjacent branches from point cloud surrounding space. The mechanism is then able avoid accurately grasp order pick it. first identified by YOLOv3 detection algorithm, space it obtained. random sample consensus algorithm RANSAC used for sphere fitting, idealized as sphere. also idealizes branch line that fitted bearing around distance between centroid constrained fitting ensure identification branch/line closest apple/sphere. best determined positional relationship straight center performance was evaluated using with various orientations on growing trees. average angle error calculated direction vector expected 11.81°, standard deviation 13.65°; 62.658% determinations erred ?10°, 85.021% ?20°. time estimating an 0.543 s. accuracy speed enabled robotic picker operate at matches human picker.