Multi-step 3D engineering models retrieval based on shape statistical information

Content-based 3D models retrieval is the key technique for reusing product design information and constructing CAD models database of enterprise. Retrieval and indexing algorithm which synthetic analyzing distance, angle of normal vector and curvature is proposed. The distance between two sampled points and the cosine value of the intersection angle of the two vertex normal vectors were employed as shape functions. The shape distribution was generated by computing the values of shape functions with sampled points. Voronoi method is used to compute the discrete mean curvature at vertices of the model mesh and probability distribution is obtained by counting these curvatures. Multi-step retrieval process for 3D engineering models is presented based on shape distribution. Experiments were conducted to evaluate the proposed method utilizing the Engineering Shape Benchmark(ESB) database.