An optimization approach to shape matching and recognition

An optimization approach to shape matching and recognition is presented. The technique is a vertex based technique that uses surface contact as a criterion for measuring similarity. Given an object (polygon O) and a reference template (polygon T), the feasible region generated by polygon T is continuously magnified (expanded or contracted) by a factor, epsilon, and polygon O is translated and rotated such that it falls inside T. The objective is to minimize the magnification factor. The problem is formulated as a nonlinear optimization problem with a linear objective function and nonlinear constraints. A search-based procedure is used to solve this problem. To ensure global optimality, the nonlinear constraints are replaced by their linear approximations.The results show that the procedure is very effective in recognizing similar objects. (C) 1998 Elsevier Science Ltd. All rights reserved.