Research On On-line Geometric Modeling Of 3d Human Body

With the rapid development of the computer technology, computer human body modeling techniques can construct high-fidelity 3D virtual human body model. The 3D virtual human body model are widely used in different kinds of industry, including area of medical, garments, television, sports, automobile an so on, that has become the focus of many researchers. With the continually progress in network technology, on-line 3D cloth fitting has become popular rapidly, and people started to buy personalized cloth customized for themselves, which is based on the construction of user's customized human body model.Based on the characteristic and requirement of on-line human body modeling, this thesis presents a method of fast human body parametric modeling. The method includes the definition of human body feature points and feature size, the definition of 3D human body model's control points, the segmentation of 3D human body model, research on shaft deformation algorithm and research on connect algorithm. By modifying OBJ file on keywords, the user-defined human body model OBJ file can be both used for human body modeling and for 3D processing softwares which support OBJ file.At the sime time, a method of photo matching modeing, that constructs human body model in higher precision with front and side contour picture, is proposed to solve problem of parametric modeing method. First, a body parts library which contains many body parts with different character is constructed, by sampling of different kinds of characteric people. Then, we research on how to extract body contour from user picture, research on how to distill body parameter from contour picture, present a method of selecting appropriate parts model from parts library automatically, and match the body parts model with user contour picture by using free-form deformation technology to make the parts model much more like the real user's body parts. Finally, the result model will be connected and be smooth by unit energy minimization method.In the field of human animation, based on skeleton subspace deformation method and characteristic of human body model, we improve on skeleton subspace deformation method by adding control ball in arthrosis, to let user view cloth fitting results facility with the reconstructed user model in common action. In the end of this thesis, some photo matching modeling results of user instance are presented to prove the feasibility of photo matching modeling algorithm of this thesis. By comparing parametric modeling results with photo matching modeling, the photo matching modeling method can construct a more realistic user model. By comparing the arm rotation result of control ball skeleton subspace deformation method with the result of traditional skeleton subspace deformation method, the former can gain a more realistic result.