STRUCTURAL METHOD FOR RECOGNIZING PARTIALLY DISTORTED CONTOUR IMAGES

This paper presents an algorithm for recognizing partially distorted t wo-dimensional objects that is invariant with respect to their size, p lacement, and orientation. The recognition is based on the composition of a semantic description of the presented object and a comparison of it with a reference description. The semantic description includes th e following list of primitives: the angular elements of the contour, s elected from the points of maximum curvature of the contour, the value s of the angles, the relative arrangement of the angles on the object, and the lengths of the line segments that form them. The results of c omputer modelling of the recognition of objects of arbitrary shape are presented. (C) 1996 The Optical Society of America.