Mesh Segmentation Via Boundary Feature Points Extraction

With the sustained development of computer technology and 3D scanning in recent years, there are many 3D models and 3D models are widely used by people. 3D mesh models are one of the most methods to represent 3D model and have been widely used in many fields, such as internet, games and manufacture industries. However, how to segment meshes into meaningful regions is useful for model understanding. At the same time, mesh segmentation is used in many applications in computer graphics, mesh segmentation is a crucial step in many practical applications, such as shape matching, shape rebuilding, skeleton extraction, texture map, mesh parameterization and so on. Therefore, mesh segmentation becomes one of important mesh manipulations in recent years.Triangular meshes are now widely used to define geometric objects in computer graphics. In this paper, we present a novel algorithm for segmenting meshes into meaningful pieces. At first, based on the minimal rule, we calculate the dihedral angle and length of the edge associated with vertices to position the vertices which probably lie on the boundary. Next, we use the clustering algorithm to clustering the boundary feature points according to distance and connectivity between points and extract parts of boundary. At last, we will get the complete boundary to segment the triangle mesh by connecting the parts of boundary.Compared with algorithms before, our algorithm only uses the dihedral angle and length of the edge when selecting features, which effectively reduce computation load and time when calculating the features. At the same time, in the process of mesh segmentation, clustering algorithms will automatically determine the number of parts which the mesh is partitioned into. At last, we assess the boundary by making use of part saliency to avoid the over-segmentation, and the effectiveness of the algorithm is proved by experiments.