基于线特征的多视点云配准算法  

A Line-based Multiview Point Cloud Registration Method

作　　者：韩一 杨文锋 徐子扬 刘莉 吉文来 HAN Yi;YANG Wenfeng;XU Ziyang;LIU Li;JI Wenlai(Guizhou Province Water Conservancy and Hydropower Survey Design&Research Institute Co.Ltd.,Guiyang 550002,China;Architectural Design and Research Institute,Nanjing Tech University,Nanjing 210009,China)

机构地区：[1]贵州省水利水电勘测设计研究院有限公司,贵阳550002 [2]南京工业大学建筑设计研究院,南京210009

出　　处：《遥感信息》2024年第5期157-161,共5页Remote Sensing Information

基　　金：贵州省水利科技经费项目(KT202111)。

摘　　要：针对多视点云配准精度低、稳定性差等问题,提出了基于线特征的多视点云配准模型(line-based multiview point cloud registration,LMR)。基于二分法的点云特征线提取方法获取多视点云特征线,利用李代数求解点云变换矩阵雅可比矩阵,构建基于非线性最小二乘优化方法的多视点云配准模型。通过配准地面激光扫描数据集验证了该方法的有效性。定量分析与结果表明,LMR提取多视点云特征线,通过全局优化多视点云位姿参数,有效提高了点云配准精度。Aiming to address the issues of low accuracy and poor stability in multiview point cloud registration,a line-based multiview point cloud registration(LMR)that utilizes line features is proposed.The feature line extraction method based on dichotomy is employed to obtain the feature lines from the multiview point cloud.Additionally,we solve the point cloud transformation matrix Jacobian matrix using Lie algebra and construct a multi-view cloud registration model based on nonlinear least squares optimization.The effectiveness of this approach is validated through registration with ground laser scanning dataset.

关 键 词：点云配准 迭代最邻近点 非线性优化 特征提取 图优化 

分 类 号：P232[天文地球—摄影测量与遥感]