基于PCA的体素化GICP点云配准算法  

Voxelized GICP Registration Algorithm Based on PCA

作　　者：徐浩鸿 付昱凯 崔世界[2] XU Haohong;FU Yukai;CUI Shijie(College of Information Science and Engineering,Northeastern University,Shenyang 110004,China;Shenyang Institute of Automation,Chinese Academy of Sciences,Shenyang 110016,China)

机构地区：[1]东北大学信息科学与工程学院,沈阳110004 [2]中国科学院沈阳自动化研究所,沈阳110016

出　　处：《组合机床与自动化加工技术》2025年第1期155-159,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金资助项目(92267201,92267205)。

摘　　要：为了满足工业中对目标工件进行三维重建的速度和精度,通过分析算法配准过程提出一种基于主成分分析(PCA)和体素化广义迭代最近点(VGICP)的点云配准策略。首先,PCA算法为精配准阶段提供良好的初始位姿,其中在进行主方向矫正时,在保证数据整体特征的基础上进行体素下采样来减少由于计算配准误差所消耗的时间,提高计算速度;其次,精配准阶段采用的VGICP算法对高度依赖最近邻搜索的GICP算法进行体素划分,使用多点分布聚合的方法,可以从较少数量的点中稳健地估计体素分布,具有较快的处理速度。基于PCA改进的VGICP算法将配准效率提高60%以上,并且优于常用配准算法,同时保持了良好的配准精度。In order to meet the speed and accuracy of three-dimensional reconstruction of target workpieces in industry,this paper proposes a point cloud registration strategy based on principal component analysis(PCA)and voxelized generalized iterative closest point(VGICP)by analyzing the algorithm registration process.First,the PCA algorithm is used to provide a good initial pose for the fine registration stage.When correcting the main direction,voxel subsampling is performed on the basis of ensuring the overall characteristics of the data to reduce the time consumed by calculating the registration error,and improve calculation speed.Secondly,the VGICP algorithm used in the fine registration stage performs voxel division on the GICP algorithm,which is highly dependent on the nearest neighbor search.A multi-point distribution aggregation method is proposed,which can robustly estimate the voxel distribution from a small number of points,with fast processing speed.The improved VGICP algorithm based on PCA improves registration efficiency by more than 60%and is better than commonly used registration algorithms while maintaining good registration accuracy.

关 键 词：体素化广义迭代最近点算法 主成分分析 点云配准 下采样 

分 类 号：TH165[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]