3D Shape Deformation Editing Based On CUDA Parallel Computing

3D shape deformation technology can be used to edit the local and global structure of3D model for obtaining nature deformation effect. It is widely utilized in the field of the product design, virtual reality and graphics processing. Based on the excellent programmable ability and highly parallel computing power of graphic process unit, this thesis adopts parallel heterogeneity mode combining GPU and CPU to execute the3D shape deformation and gain real-time and realistic shape deformation.An interactive3D shape deformation method is researched. Using edge collapse mesh simplification algorithm, the reduced mesh is extracted from the original shape and retains the characteristics as much as possible. One can set the fixed and the moved constraints for some points in the reduced mesh interactively, and drag the moved handle to a preferred location. By specifying affine transformations for the rest points in the reduced mesh, the matrix parameter elements are computed by Gauss-Newton iteration and then the coordinates of all the vertexes in the original shape are calculated and updated. So we can obtain the deformed3D shape via rendering.Based on compute unified device architecture, the acceleration computing is researched for3D shape deformation using programmable GPU. As the treatment of the vertexes in the calculating process are independent, the rotation matrix and translation vector are able to be computed in parallel. For the same updating coordinate instruction, multiple threads run at the same time, and the vertexes are mapped to the new positions. By dividing deformation task, minimizing data transfers between host and device, as well as optimizing device memory accessing, the efficiency of parallel computing is improved.3D shape deformation experiments is implemented for numbers of3D models, the results show that the realistic effect can be generated and several dozen times of speedup ratio can be achieved comparing with the serial implementation mode. The tradeoff between the effect and the time consumption is realized for the complex3D shape deformation.Based on the deformation algorithm and parallel computing mode, shape Editing architecture is designed. By dividing structural hierarchy and implement steps, a fast3D shape deformation system is realized using Visual Studio and CUDA platform.