Efficient shape matching through model-based shape recognition

Abstract

An efficient shape matching method for shape recognition is proposed. It first uses a polygonal approximation technique to describe a shape. Then, using the proposed matching algorithm (based on some polygonal attributes), the dissimilarity measure between an input shape and its reference model is obtained. The input shape can be classified into the class (model) which has the minimum dissimilarity measure with it. In addition, it can provide the best matching orientation information for the input shape and its model, so that further applications, such as automated inspection and assembly, can be effectively performed. The proposed shape matching is invariant to translation, rotation, and scale changes of a shape. Experiments using noisy shapes and real images are performed, and the recognition results demonstrate that the proposed method is very effective.