Abstract

With the development of graphic accelerated hardware and 3D modeling tools, 3D models will be as prevalent as other multimedia data in the future. Thus, effective content-based 3D model retrieval systems are required for emerging needs. Many 3D model retrieval methods have been proposed in recent years. Shape distributions showed superiority over others due to rotation invariance and ease of computation, but the discriminative accuracy is limited for any loss in information. Fuzzy Shape Distributions (FSD) is proposed to improve the retrieval performance of distribution-based methods. First, two improved shape distributions are presented by using the concentric partition and symmetrical partition for 3D models. Second, the two enhanced descriptors are combined with a fuzzy weighted procedure. Experimental results show that the proposed FSD can achieve better retrieval performance.

Key words: 3D model retrieval, fuzzy weighted, shape distribution, sequential quadratic programming, content-based.