Skeleton Matching Based On Clustering

The skeleton is a one-dimensional representation of three-dimensional geometry, and it has been widely used in computer graphics and visualization. There is a lot of study of the skeleton, and we have made a compelling application of the results in animations, deformation, shape registration, shape retrieval, object recognition, matching skeleton, surface segmentation, and so on. A commonly used method in the shape retrieval is to extract the skeleton of the object, and then retrieve the similar graphics from the graphics library by considered the match degree of the skeleton of the object. So the key step in the shape retrieval is effective skeleton matching. Surface segmentation can be done by using the skeleton, especially for the case that the two deformation models need to be divided into the same number of surface patches and a continuous mapping need to be created between the surface patches of the two models. In this paper we propose a new method for skeleton matching based on this application.In this paper, we introduce a few basic methods for skeleton extraction and some criteria for evaluating the quality of the skeletons firstly. Then, we briefly introduce some shape matching methods. The main work in this paper is that we propose a new method for graph matching based on clustering. Continuous sampling the moving objects, our goal is that we divide the skeletons of the adjacent two frames into some corresponding smaller parts, and they have the same number. The specific approach is to extract the key points from the skeletons, which are divided into the endpoints and cross points of the skeleton, and the matching of them should be done separately. Firstly, cluster the endpoints of the skeletons, and match the endpoints based on the skeleton path between the clustering, and then determine the matching relationship of the junction points by using the similar method.