3D Mesh Hierarchical Segmentation And Skeletal Extraction

In recent years, animation has become a hot research topic, 3D model has also appeared in a number of industries.3D model has become an integral part of development of the industry. Extensive using of 3D model, people are triggered to concern the treatment of 3D model, 3D model of segmentation and skeleton extraction is one of the important research. 3D grid model segmentation technology is widely used in many fields, such as modeling, 3D model deformation, and mesh simplification. 3D model skeleton extraction is an indispensable component of animation. Currently, there are many grid segmentation algorithm has been proposed, such as the watershed algorithm, clustering algorithm, region growing algorithm and so on.This paper presents a novel 3D model segmentation and skeleton extraction algorithm, semi-supervised 3D model is based on semi-supervised segmentation level thought, using of K-means clustering and Gaussian curvature to extract the skeleton. This segmentation algorithm is divided into three parts: extraction of the features patches of the model, pre-split and post split. The features patches of the 3D model which based on multi-Dimensional Scaling method, originally defined from the European space of the original model to MDS space for getting a new model. We get the pre-segmentation result by K-means clustering; the model first for a rough segmentation algorithm will help improve overall operating efficiency. In the pre-split basis, using the idea of regional growth and reading zonal model, which is the discrete Gaussian curvature smaller place. By the model segmentation the boundary more meaningful segmentation result is better.First, the model according to the segmentation algorithm is divided into the designated section, then each partition in the model module to feature patches as seed points to gradually read the neighborhood with the regional growth, promoting the reading level of the partition block; calculated for each level with the center forward, in turn connected center shall be the skeleton line with blocks, each separated by branches of the module frame, the next step under the partition between the adjacent part of nature, between the adjacent split pieces skeleton linking, that is, the overall model framework. Experimental results show that the algorithm is correct and effective.