基于混合优化的快速隐式曲面采样方法  

Rapid Sampling on Implicit Surfaces Via Hybrid Optimization

作　　者：李伟涛[1] 周元峰[1] 高珊珊[2] 张彩明[1,2] 

机构地区：[1]山东大学计算机科学与技术学院,济南250101 [2]山东财经大学山东省数字媒体技术重点实验室,济南250014

出　　处：《计算机学报》2014年第3期593-601,共9页Chinese Journal of Computers

基　　金：国家自然科学基金(61020106001;61202148;61373078);山东省优秀中青年科学家奖励基金(BS2013DX041;BS2013DX048);山东大学自主创新基金(2012TB013)资助

摘　　要：该文提出了一种新的隐式曲面快速采样方法,该方法首先提出了一种新的采样点互斥能量目标函数,基于该目标函数,通过一种混合优化方法来求解采样点的分布.第1步为采样点的局部优化,通过对采样点移动速度的控制参数调整,避免了大量Hessian矩阵的求逆操作,使得采样点能够根据互斥半径快速覆盖整个隐式曲面,得到初始采样点集;第2步为采样点的全局优化,采用L-BFGS方法对所有采样点进行优化,得到最终的高质量采样结果.通过实验表明,新方法的采样速度大大提高,并能够获得较好的隐式曲面采样点分布.A new method for implicit surfaces sampling is presented in this paper. At first, a new objective function of repulsive energy is given for constraining sampling points distributed on the surface uniformly. Based on this new objective function, the distribution of sampling points can be solved by a hybrid optimization. The first step is a local optimization of sampling points through a parameter to control the velocity of a sampling point for avoiding inverse matrices com- putation of Hessian, so a set of sampling points can cover given implicit surface rapidly. The second step is global optimization of sampling points. L-BFGS method is used in this step. We can have ideal sampling distribution via these two steps. Experimental results show that the new method can sample fast on implicit surfaces, and the points＇ distribution is also satisfactory.

关 键 词：隐式曲面 动态采样 优化 目标函数中图法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]