A visual imitation learning algorithm for the selection of robots’ grasping points