LiDAR点云辅助的无人机影像快速高质正射纠正方法  

A Rapid and High-quality Orthorectification Method of UAV Image Assisted by LiDAR Point Cloud

作　　者：吴宇峰 李大军[1] 刘华[1] 黄婷婷 李博宇 WU Yufeng;LI Dajun;LIU Hua;HUANG Tingting;LI Boyu(School of Surveying and Geoinformation Engineering,East China University of Technology,330013,Nanchang,PRC;China Railway Water Resources and Hydropower Planning and Design Group Co.,Ltd.,332029,Nanchang,PRC)

机构地区：[1]东华理工大学测绘与空间信息工程学院,南昌330013 [2]中铁水利水电规划设计集团有限公司,南昌332029

出　　处：《江西科学》2025年第2期356-362,共7页Jiangxi Science

基　　金：江西省科技重点研究计划项目(20223BBE51030)。

摘　　要：针对传统基于密集匹配生成DSM(Digital Surface Model,DSM)的无人机正射影像生成方法虽能确保高精度,但有效率低下的问题,提出了一种LiDAR点云辅助的无人机影像纠正方法。首先利用影像三维特征点与LiDAR点云的同名点作为控制约束;然后滤波LiDAR点云数据以获得高精度DEM(Digital Elevation Model,DEM);最后利用DEM对影像进行正射纠正。在3个不同场景下,分别采用本算法、商业软件Pix4D以及基于空三稀疏点云的正射影像方法进行对比实验。结果表明,本算法所生成的正射影像细节表现上略优于另外2种方法;同时,相较于商业软件Pix4D本算法在效率上提升了1个数量级,且达到了相当的几何精度,平面误差的均值为0.023~0.029 m;平面均方根误差为0.024~0.031 m,因此LiDAR点云辅助的无人机影像纠正方法能够快速生成高精度的数字正射影像。To address the issue of low efficiency in traditional UAV orthophoto generation methods based on dense matching for DSM(Digital Surface Model)generation,which ensures high accuracy but is time-consuming,a LiDAR point cloud-assisted UAV image correction method was proposed.First,3D feature points from the images and corresponding points from the LiDAR point cloud were used as control constraints.Then,the LiDAR point cloud data were filtered to obtain a high-precision DEM(Digital Elevation Model).Finally,the DEM was used to orthorectify the images.Comparative experiments were conducted in three different scenarios,using the proposed algorithm,commercial software Pix4D,and an orthophoto generation method based on sparse point clouds from aerial triangulation.The results indicate that the orthophotos generated by the proposed algorithm exhibit slight better detail representation than the other two methods.Additionally,the proposed algorithm improves efficiency by an order of magnitude compared to Pix4D while achieving comparable geometric accuracy,with a mean planar error ranging from 0.023 to 0.029 meters and the root mean square error(RMSE)of 0.024 to 0.031 meters.Therefore,the LiDAR point cloud-assisted UAV image correction method can quickly generate high-precision digital orthophotos.

关 键 词：数据融合 机载激光扫描 LIDAR点云 POS数据 正射纠正 稀疏点云 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]