基于改进的点云配准算法的室内2D建图方法  

Indoor 2D Mapping Method Based on Improved Point Cloud Registration

作　　者：吴雪静 苏杰 宋席发 李杨[1] 韩笑 

机构地区：[1]北华航天工业学院,河北廊坊065000

出　　处：《工业控制计算机》2025年第1期100-102,共3页Industrial Control Computer

基　　金：北华航天工业学院博士基金项目(BKY-2020-19、BKY-2021-14);北华航天工业学院科研基金项目(CXPT-2021-02)。

摘　　要：利用单线激光雷达的二维点云数据构建地图是一种常见的室内环境建图方法。对于传统的ICP算法在点云配准过程中存在初始对齐敏感而容易陷入局部最优解的问题,采用分步配准与改进的点对点ICP算法,对点云数据的匹配进行改进,有效地解决旋转角度较大或偏移距离较远时的精确匹配问题。针对单线激光雷达扫描得到的两帧点云数据,首先分别进行去噪、滤波等预处理操作,然后使用PCA粗配准算法和改进的点对点ICP算法对两帧数据进行融合。最后,通过RANSAC算法对二维点云进行分割与拟合,并完成建图。实验结果表明,该方法在保证精度的同时,能大幅提高建图的效率。Constructing maps using two-dimensional point cloud data from single-line lidar is a common method for indoor environment mapping.However,the traditional ICP algorithm is sensitive to initial alignment and easily falls into local optima during point cloud registration.This paper proposes an improved point-to-point ICP algorithm combined with stepwise registration to address the issue of accurate matching in cases of large rotation angles or significant displacement distances.For two frames of point cloud data obtained from single-line lidar scans,this paper first performs pre-processing operations such as denoising and filtering,and uses the PCA coarse registration algorithm and the improved point-to-point ICP algorithm for data fusion.Finally,the two-dimensional point cloud is segmented and fitted using the RANSAC algorithm to complete the mapping process.Experimental results demonstrate that our method significantly improves mapping efficiency while maintaining accuracy.

关 键 词：单线激光雷达 ICP算法 地图融合 室内建图 

分 类 号：TN958.98[电子电信—信号与信息处理]