On the recognition of industrial parts

Abstract

The authors present an algorithm for representing two-dimensional shapes that may be partially occluded or overlapped. Based on the approximated polygon, extended features that encompass neighborhoods of significant corners are defined. Two hypothetic angularities are first established to create hypothetic space. An analytic-circular-tag (ACT) which groups a specified number of consecutive hypothetic points in the sequence is used to provide a new scheme for matching technique. A hypothesis-generation-testing technique is used for contour matching. The experimental results demonstrate that the algorithms can be used to match complex industrial parts.<>