Method of 3D model retrieval by the spatial distributions of components

3D shape analysis is a key problem for 3D model retrieval. In this paper, a shape signature describing the spatial distributions of 3D model's components is proposed. The primary motivation of the approach is to emphasize the object's structural attributes due to human vision theories. Firstly, a model is decomposed into several meaningful components, each of which is represented by a sub-mesh, then the component's mass center and percentage of the surface area are combined to form a component feature, and finally a set of component features satisfing the specified conditions are used as the shape signature of a 3D model. Based the signature, an approach of 3D model retrieval is presented. It performs well despite the presence of different resolutions and connectivity, and is fast when measuring shape similarity. Experimental results demonstrate the ability of the proposed signature to retrieval 3D modes.