Three Dimensional Human Body Model Deformation Based On Semantics

3D human body model is an indispensable element in films and computer games. It also has many applications in medical equipment, clothing design, image processing, etc. Human body model that is obtained through scanning cannot be used directly. As the scene may change, moderating the size of the current model is the priority of model manipulation. Meanwhile, the operation difficulty of model manipulation, the practicability of deformation and its effect also require consideration. With these questions in mind, we’ve learned and referenced a lot of deformation methods and brought in semantics driving model deformation method on the basis of the existing models and methods. By measuring the bone joint point position of the model, we’ve got semantics values like height, length of legs and height of shoulders in order to adjust the related semantics value of model, finally got the new model after adjustment.The main purpose of this thesis is to obtain the semantics parameters by measuring the model and further drive model deformation by adjusting these parameters.The first thing to do is obtaining semantics of model. The first step is to measure height, length of legs and height of shoulders of model by extracting the coordinates of joint point. Then calculate waist and bust circumference of the joint point of the model in order to get fitting function of waist, bust circumference and weight on the basis of MPI laboratory model library. By verifying the error of fitting functions within the scope of permit, we import model library of Princeton university in the system to calculate the corresponding semantics values.The second step is to adjust semantics values and to drive model deformation. Use PCA dimension reduction to obtain the model parameter values needed for the deformation of model then map the model to PCA space. Modify semantics values of model to drive model deformation.On the basis of this procedure, we designed a software platform for studying3D human model driven by semantics to experiment. By comparing the experiment results, it turns out to be a better deformation result through model deformation driven by semantics.