A skeleton-retrieving approach for the recognition of punch shapes

Abstract

Automatic punch shape recognition is a key task in the design automation of progressive dies. Many researchers have proposed and used different algorithms to represent global shape information of 2D objects by their skeletons. This paper briefly examines the current state of research and application of shape skeletons for image processing and engineering. It is found that Wu et al.'s simplified skeleton is most suited for the representation and recognition of punch shapes in a die design automation system. This paper explains how the enhanced simplified line skeleton, a modification of Wu et al.'s simplified skeleton is able to represent local shape information in addition to the global shape information. As such, the enhanced simplified line skeleton has sufficient features for the recognition of a set of 78 basic and special standard punch shapes provided in a commercial die component catalogue. A hierarchical search method for punch shape matching is also explained. This method is more efficient than a geometrical feature-based approached originally used to develop a die design automation system at the National University of Singapore. © 1997 Elsevier Science B.V.