基于随机采样一致性算法的激光点云大数据自动配准技术  

Automatic registration technology of laser point cloud big data based on random sampling consistency algorithm

作　　者：潘莹 刘玉丽 王君 PAN Ying;LIU Yuli;WANG Jun(School of Information Engineering,Herbin Univercity,Harbin 150086,China)

机构地区：[1]哈尔滨学院信息工程学院,哈尔滨150086

出　　处：《激光杂志》2023年第5期128-132,共5页Laser Journal

基　　金：黑龙江省高等教育教学改革研究项目(No.SJGY20200419)。

摘　　要：激光点云大数据中存在冗余数据,导致其大数据自动配准过程中存在配准速率慢、精度差等问题,为此提出基于随机采样一致性算法的激光点云大数据自动配准技术。将初始半径和最小邻域拟作自定义变量,使用密度聚类法剔除冗余激光点云数据,运用双边滤波平滑点云信息;分析不同角度下两组激光点云数据的刚体变换关联,获取对应且不共线的采样点,将模型点配准率作为激光点云采样数据和真实数据一致性衡量指标,完成激光点云大数据初始配准;通过迭代最近法找出两个点云集合的最近点,利用M-估计方法优化误差惩罚函数,实现激光点云大数据自动配准。仿真结果证明,应用所提方法得到在未引入噪声时的激光点云数据配准重合度为0.97,噪声环境下的点云配准重合度为0.96,配准平均耗时为59.6 s。There are redundant data in the big data of laser point cloud,which leads to the problems of slow regis-tration rate and poor accuracy in the process of automatic registration of big data.To this end,an automatic registration of big data of laser point cloud based on random sampling consistency algorithm is proposed.quasi-technical.The ini-tial radius and the minimum neighborhood are assumed to be custom variables,the redundant laser point cloud data is eliminated by the density clustering method,and the point cloud information is smoothed by bilateral filtering;the rigid body transformation correlation of the two sets of laser point cloud data at different angles is analyzed to obtain the cor-responding For the sampling points that are not collinear,the model point registration rate is used as a measure of the consistency between the laser point cloud sampling data and the real data,and the initial registration of the laser point cloud big data is completed.At the nearest point,the M-estimation method is used to optimize the error penalty func-tion to realize automatic registration of laser point cloud big data.The simulation results show that the registration coin-cidence degree of the laser point cloud data is 0.97 when no noise is introduced,and the registration coincidence de-gree of the point cloud in the noise environment is 0.96,and the average registration time is 59.6 s.

关 键 词：随机采样一致性 激光点云 数据配准 ICP方法 双边滤波 

分 类 号：TN958[电子电信—信号与信息处理]