【摘要】 针对大规模城市场景下同名线对精度不足导致点云配准不准确的问题，提出了基于线特征的大规模城市场景点云配准算法。首先，该方法对机载点云和车载点云中的建筑物进行3D特征线的检测和筛选，并将特征线分为横向特征线和纵向特征线两组；其次，基于线段豪斯多夫距离（Line segment Hausdorff Distance, LHD）进行初始3D同名线对的搜索；然后，通过平面拟合与相交的方法对精度不足的纵向同名线对进行优化，并设计不同组合的同名线对配准方案；最后，利用RLMR-FMII₂配准算法实现不同配准方案的点云精确配准，得出最优的配准方案。采用实测点云进行实验，得出基于优化后的纵向同名线对的配准方案为最优方案，实验结果表明该算法能够实现大规模点云的精确配准。更多还原

【Abstract】 Aiming at the problem of inaccurate point cloud registration due to insufficient accuracy of corresponding line pairs in large-scale urban scenes, a point cloud registration algorithm for large-scale urban scenes based on line features is proposed. Firstly, the method detects and screens the 3D feature lines of buildings in airborne point cloud and vehicle-borne point cloud, and the feature lines are divided into two groups: horizontal feature lines and vertical feature lines; secondly, the initial 3D corresponding line pairs is searched based on the line segment Hausdorff distance（LHD）; then, the corresponding line pairs with insufficient accuracy are optimized by plane fitting and intersection methods, and different combinations of the corresponding line pair registration schemes are designed; finally, the existing RLMR-FMII₂ registration algorithm is used to achieve accurate registration of point clouds with different registration schemes, and the optimal registration scheme is obtained. Using the measured point cloud to conduct experiments, it is concluded that the registration scheme based on the optimized vertical corresponding line pairs is the optimal scheme. The experimental results show that the algorithm can achieve accurate registration of large-scale point clouds.更多还原