The Application Research of Robot Vision Technique on Piston´s Online Recognition and Selection

A step dynamic object positioning method of robot eye-in-hand vision is proposed. An experimental platform is developed in which the CCD plane is in parallel with worktable plane, images acquired by camera at several fixed positions and target is positioned at the center of the image finally. During the recognition procession of piston type, BP neural network is used to recognize the characters on piston´s surfaces, piston type is classified according to the characters. By the combination of recognized characters and piston´s position information, the end-executor is led to select the piston accurately.