近似刚性的快速点云变形算法  

As-rigid-as-possible Efficient Deformation Algorithm for Point Clouds

作　　者：赵勇[1] 刘光鹏[1] 彭群生[2] 

机构地区：[1]中国海洋大学数学科学学院,青岛266100 [2]浙江大学CAD&CG国家重点实验室,杭州310058

出　　处：《计算机辅助设计与图形学学报》2013年第7期955-962,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家"九七三"重点基础研究发展计划项目(2009CB320802);山东省优秀中青年科学家科研奖励基金(BS2012DX043);中央高校基本科研业务费青年教师专项基金项目(201313005)

摘　　要：针对大规模的复杂点云模型,提出了一种非线性变形算法.通过保持局部邻域的刚性使得点云模型进行近似刚性的变形,在大尺度变形时有效地保持几何细节和体积;为了保证求解过程的收敛性,对原始点云进行聚类,快速生成其简化点云,进而在稀疏的简化点云上完成变形,并将该变形作用给原始点云以得到合理的初始结果.此外,还给出一种动态重采样方法,以消除变形造成的冗余点和裂缝.实验结果和对比数据表明,文中算法简单高效,能够防止几何细节的扭曲和明显的体积变化,获得了令人满意的变形效果.It is a challenging problem to efficiently deform complex large point clouds in computer graphics. In this paper, a novel non-linear algorithm is presented to achieve as-rigid-as-possible deformation through constraining the rigidity of the local neighborhoods, thus preserving both detail and volume under large deformations. To guarantee the convergence of the numerical solver, the original point clouds is efficiently simplified by clustering, then the deformation is performed on the simplified point clouds and transferred to the original point clouds to obtain the initial result. Furthermore, a dynamic resampling method is introduced to eliminate the redundant points and cracks that often occur during deformations. The experiment data demonstrate that our algorithm is easy to use, efficient and effective in preventing detail distortions and volume changes.

关 键 词：点云模型 近似刚性的变形 快速收敛 动态重采样 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]