Abstract: Segmenting and recognizing the feature of CAD meshes enhance the efficiency of model reusing and editing in the design of complex mechanical products. An algorithm for segmenting CAD meshes based on the Gaussian map is presented. The Gaussian map of the triangular faces is constructed and the adjacent connectivity of the point in the Gaussian sphere is also created. The points are classified into several patches by k-means method, and each patch is separated into sub-patches by the region growing method. The small patch is merged into adjacent patch, and then the planar region, the cylindrical region, the conical region, the spherical region, and the toroidal region are recognized and merged into the similar regions based on the adjacent matrix of each patch. The empirical results show that the proposed algorithm is efficient and robust in clustering and recognizing CAD models of complex mechanical product.

Key words: CAD mesh segmentation, Gaussian map, k-means normal clustering