基于自由曲面的点云配准算法  

Point cloud registration algorithm based on free-form surface

作　　者：卢章平[1] 郑航[1] 沙春发[1] 李明珠[1] 

机构地区：[1]江苏大学机械工程学院,江苏镇江212013

出　　处：《江苏大学学报（自然科学版）》2015年第3期319-323,共5页Journal of Jiangsu University：Natural Science Edition

基　　金：高等学校博士学科点专项科研基金资助项目(20113227110007)

摘　　要：为了提高自由曲面工件的配准效率,提出了一种基于共面4点集的RANSAC初始配准算法和改进的迭代最近点(ICP)精确配准算法相结合的2步配准方法.首先,在基于RANSAC算法的机制上,通过点间距离和比例关系寻找2片点云的共面4点集,利用共面4点集这一不变量来约束RANSAC算法提取的样本,使点云经过初始配准后得到一个较好的初始位置;然后在基于原始ICP算法的基础上作出相应的改进,对点云初配结果进行优化,使得点云之间的配准误差达到最小,以实现点云的精确配准;最后,对2组简单工件的CAD曲面点云模型进行配准仿真.结果表明:该算法相对于传统ICP算法运行时间减少48%,精度提高56%,能够满足配准要求.To improve the registration efficiency of free-form surface work piece,according to RANSAC preliminary registration algorithm,a new two-step registration method of free-form surface was proposed based on coplanar 4-point sets algorithm and improved iterative closest point（ ICP） exact registration algorithm. The coplanar 4-point sets were searched through the distance and proportional relationship among points based on RANSAC. The invariants of coplanar 4-point sets were used to constraint the sample extracted by RANSAC algorithm,and the point cloud was marked to get a good initial position after preliminary registration. The method was improved based on original ICP algorithm to optimize the result of preliminary registration. The registration error was minimized to realize the exact registration. The registration simulation of CAD surface cloud models of two simple work pieces was completed. The results show that compared to traditional ICP algorithm,the running time of the proposed algorithm was reduced by48% with increased accuracy by 56%.

关 键 词：点云配准 自由曲面 共面4点集 RANSAC算法 迭代最近点算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]