(A) machine vision system that recognizes position and orientation of parts

Abstract

A machine vision system is described which can recognize the position and orientation of isolated parts that can be found in assembly belt line or lathe machining process. The system reduces grey-scale image data to a set of compact edges which has the appearance of digital straight line. The outer-most boundary is represented by chain codes and centers of inner holes are found. Corners of image object are found, as features, from these chains and compared with the model which is readily formed by human operator with digitizer. In matching process, a straight line approximation of an object is developed in which corners of models are linked as a set of ordered vector list and they are compared with the model vectors. The images are obtained in a well contrasted condition where an object is far brighter than the background and has no reflection. The system has been programmed in PL/I on DEC VAX-11/780 computer and experiments have been conducted with a reasonable success.