Your conditions: 龚随
  • 基于低精度布料采样的多精度布料构建方法

    Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》

    Abstract: In order to take into account the fidelity and computational efficiency of cloth simulation, this paper presented a construction method of multi-resolution cloth based on low-resolution cloth sampling. Firstly, through sampling a simulation example of low-resolution cloth movement several times, this paper obtained the average deformation degree, which represented by the average deformation degree of the vertex and the edge collision mark. Then, according to the results of sampling, it divided the region of low-resolution into high deformation area, middle deformation area and low deformation area. Next, it used the improved adaptive subdivision algorithm to refine the three deformation areas in different extent, so as to construct the multi-resolution geometric model. Finally, through the definition of the mass and the spring coefficient of the cloth, it obtained the multi-resolution physical model. This paper reduces multi-resolution grid number and improves computational efficiency relative to the high-resolution grid. It also improves the fidelity of cloth simulation compare to the low-resolution grid.

  • 基于椭球拟合的人体—服装碰撞检测方法

    Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》

    Abstract: To realize the rapid collision detection between cloth and human body in cloth simulation, a collision detection method based on ellipsoid fitting is proposed. Firstly, the geometric distance isoline was used as the basic data. And the model feature points was extracted by combining the linear relationship between body size and height with the structural features of the human body, which realized the semantic segmentation of the model. Secondly, the average of the radial distance was used to represent the fitting error between an ellipsoid and the model. And the number of cluster centers was gradually increased by bisecting K-means clustering algorithm based on pruning optimization. It could realize the rapid clustering of human model and generate a series of minimum ellipsoids. Finally, cloth grid points was detected with the ellipsoids instead of the model. Simulation experiments show that this method not only realizes the fast fitting of the human model, but also effectively improves the computational efficiency of collision detection.