基于局部点对特征与ICP的粗-精点云配准算法  

Coarse-to-fine registration algorithm based on local point pair feature and iterative closest point

作　　者：岳晓峰[1] 刘泽园 朱娟 田云胜 YUE Xiaofeng;LIU Zeyuan;ZHU Juan;TIAN Yunsheng(School of Mechatronic Engineering,Changchun University of Technology,Changchun 130012,China)

机构地区：[1]长春工业大学机电工程学院,吉林长春130012

出　　处：《长春工业大学学报》2022年第4期306-314,共9页Journal of Changchun University of Technology

基　　金：吉林省发改委项目(2020C018-3)。

摘　　要：三维点云配准在机器视觉与智能检测领域中有着广泛应用,也是近年来工业机器人自动拾取的基础技术。针对传统迭代最近点(ICP)配准算法存在收敛速度慢、对点云初始位置要求高等问题,提出一种基于局部点对特征与ICP的粗-精点云配准算法。配准实验表明,该算法在最优的情况下可使ICP算法的迭代次数减少73%,有效解决ICP算法鲁棒性差、效率低等问题。3D point cloud registration is widely used in the field of machine vision and intelligent detection.It is also the basic technology of industrial robot automatic picking in recent years.Aiming at the problems of slow convergence and high requirement for initial position of point cloud in traditional iterative closest point registration algorithm,a coarse-fine point cloud registration algorithm based on local point pair feature and iterative closest point algorithm is proposed.Registration experiments show that the algorithm can reduce the number of iterations of ICP algorithm by 73%in the optimal case,which can effectively solve the problems of poor robustness and low efficiency of ICP algorithm.

关 键 词：点云配准 点对特征 迭代最近点算法 特征匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]