Abstract: Although the classical global shape recognition algorithm is efficient, it is not good enough to deal with the deformation. The local shape recognition algorithm has a good retrieval rate, however, the discriminability still needs improvement. To solve these problems, a fusion framework based on the classification of characteristic points is proposed, which not only takes the advantages of global and local shape recognition algorithms, but also makes up for the lacks of the two algorithms. Some classic shape recognition algorithms build the shape feature histogram by extracting the characteristic points. Based on this, these points are further classified that different shape recognition algorithms are applied to different kinds of points. The matching results are then fused to make full used of the advantages of the global and local shape feature descriptors. Experiment results show that the proposed framework can effectively combine different algorithms to achieve the shape recognition and get better results.

Key words: computer application, shape recognition, characteristic point, shape feature histogram, feature fusion