Abstract:Aiming at the problems of fixed workpiece position and low grasping efficiency caused by robot teaching programming method, this paper studies the application of neural network in robot vision recognition and grasping planning, establishes a vision guidance scheme, develops a vision recognition system through YOLOV5 neural network model, identifies the types of objects, and obtains the coordinates of the positioning points of the objects to be grasped. The principle of robot six-point hand-eye calibration is put forward and the calibration experiment is carried out. The positioning method for objects with circular or rectangular top view is put forward. Finally, 180 grabbing experiments were carried out for three kinds of objects, and the overall average success rate of the experiments was about 92.8%, which verified the possibility of practical application of the vision recognition and grabbing robot system and effectively improved the grabbing efficiency.