Research And Implementation Of Skeleton Extraction Method Of 3D Human Body Point Cloud Model

Three-dimensional model skeleton has been widely used in three-dimensional animation,three-dimensional model retrieval and matching,medical imaging,computer simulation,mechanical engineering,virtual reality and many other fields.In this paper,we take 3D point cloud as the study object and do research on the method of 3D skeleton extraction.The main research work is as follows:(1)We establish an initial skeleton method of point cloud model which is based on the slice.Before slicing the point cloud model,we calculate the tilt of the model using PCA and the minimum bounding box principle and rotate the whole model vertically.Then,we cut and project the model according to the direction of the main trend combine with a certain step size and the direction of the main trend after rotation.And then,we use the grid growth algorithm to divide region for each projection layer and get the center of each area of the projection layer.Last,we mark each regional center point again to generate the model skeleton point.(2)We present a model initial skeleton generation method based on the initial skeleton point.The method first determines the range of the points according to the layer and removes the skeleton points located outside the model.Then connect the same part of the model skeleton point according to the projection layer number and mark number at the skeleton point.And form the skeleton lines of all the components of the model.Last connect the body and limb branches at the skeleton point to form the initial skeleton line according to the human body model topology.(3)We designed a two-stage approach to skeleton optimization.The two phases include coarse optimization and fine optimization adjustments.For coarse optimization adjustments,we merge the internal error bifurcation points due to noise and other causes on the basis of the initial skeleton.And we move the skeleton lines outside the model to the interior of the model based on the interpolation skeleton points at the same time.For the fine optimization adjustments,we adjust the branch further according to the coarse adjustment of the skeleton using the inserting interpolation point method.That ultimately makes the model branch closer to the model's axis position.Under the premise of guaranteeing the basic topological relation of human model,using the methods of this paper the axial and smoothness of the skeleton of the human body model are more accurate and do not need to be beforehand denoised and meshed in the model,which increases the robustness of the method.