Detection Of The Boundary Points D Body Model Segmentation

As the continuous development of three-dimensional mesh model reconstruction and visualization techniques, as well as high precision and advanced scanning technology, the amount of data and complexity of the3D body model are also increasing, in many cases much higher than the current computer graphics processing capabilities. As the model segmentation can largely reduce the complexity of the algorithm of the mesh, a number of mesh-based segmentation algorithm emerged, most of these algorithms using the human body model segmentation as a preliminary stage,the three-dimensional model segmentation of human body has become a vital part of most human mesh algorithm. The model segmentation of human body have great research value in the skeleton extraction, mesh deformation, boundary detection and texture mapping and so on.The research of this paper based on the boundary point detection of three-dimensional human body model. We adopted the method of learning and trained a three-dimensional human body model boundary point detection of AdaBoost classifier, achieve the boundary point detection of three-dimensional human body model,and prepare for the model segmentation of human body.The paper first calculate the five characteristic points at the end of the human body model by the geodesic distance and Dijsktra algorithm shortest path, then mark the boundary point and the non-boundary point that used for training AdaBoost classifier based on the geodesic distance relation with the ratio between the each part of human body. Second, we calculate the six characteristic value about boundary point and non-boundary point according to the selection characteristic of maximum curvature, minimum curvature, gauss curvature, mean curvature, the shape index and curvedness. We constructed an AdaBoost training dataset by the six characteristic value of boundary point and non-boundary point that calculated before and trained AdaBoost classifier, then we take advantage of the classifier which trained to detect boundary points of three-dimensional human body model. The result of experiment indicate that the method of boundary point detection about human body model is feasible, boundary point detection based on three-dimensional human body model segmentation has certain practical significance in a three-dimensional human body model segmentation study.