A PARTICLE FILTER APPROACH TO LEARNING PARTIAL SHAPE CORRESPONDENCES

摘要

This paper describes a Particle Filter approach to solve the partial shape correspondence problem. The shapes are represented by ordered sequences of local features along their polygonal boundary. Learning domain knowledge which specifies additional global constraints, the Particle Filter system finds locally and globally consistent correspondences between similar shape parts. Experiments using standard alignment techniques based on the given correspondence relationships, demonstrate the advantages of this approach, outperforming related approaches in partial shape matching.