It is still a completely new and challenging task to register extensive, enormous and sparse satellite light detection and ranging (LiDAR) point clouds. Aimed at this problem, this study provides a ridgeline-based terrain co-registration method in preparation for satellite LiDAR point clouds in rough areas. This method has several merits: (1) only ridgelines are extracted as neighbor information for feature description and their intersections are extracted as keypoints, which can greatly reduce the number of points for subsequent processing, and extracted keypoints is of high repeatability and distinctiveness; (2) a new local-reference frame (LRF) construction method is designed by combining both three dimensional (3D) coordinate and normal vector covariance matrices, which effectively improves its direction consistency; (3) a minimum cost–maximum flow (MCMF) graph-matching strategy is adopted to maximize similarity sum in a sparse-matching graph. It can avoid the problem of "many-to-many" and "one to many" caused by traditional matching strategies; (4) a transformation matrix-based clustering is adopted with a least square (LS)-based registration, where mismatches are eliminated and correct pairs are fully participated in optimal parameters evaluation to improve its stability. Experiments on simulated satellite LiDAR point clouds show that this method can effectively remove mismatches and estimate optimal parameters with high accuracy, especially in rough areas.

中文翻译：

---

基于脊线的地形共配准，用于粗糙区域中的卫星LiDAR点云

记录大量，稀疏的卫星光检测和测距（LiDAR）点云仍然是一项全新的挑战性任务。针对这一问题，本研究提供了一种基于脊线的地形共配准方法，为粗糙地区的卫星LiDAR点云做准备。该方法有以下优点：（1）仅提取山脊线作为邻域信息进行特征描述，并提取其交点作为关键点，可以大大减少后续处理的点数，提取的关键点具有较高的重复性和鲜明性；（2）结合三维（3D）坐标和法向矢量协方差矩阵，设计了一种新的局部参考系（LRF）构造方法，有效地提高了其方向一致性。（3）采用最小成本-最大流量（MCMF）图匹配策略来最大化稀疏匹配图中的相似度和。它可以避免传统匹配策略引起的“多对多”和“一对多”问题。（4）采用基于最小二乘（LS）的配准的基于变换矩阵的聚类，其中消除了不匹配，并且正确对完全参与了最佳参数评估，以提高其稳定性。在模拟卫星LiDAR点云上的实验表明，该方法可以有效地消除失配并以高精度估计最佳参数，尤其是在崎rough地区。由传统的匹配策略引起；（4）采用基于最小二乘（LS）配准的基于变换矩阵的聚类，其中消除了不匹配，并且正确对完全参与了最佳参数评估，以提高其稳定性。在模拟卫星LiDAR点云上的实验表明，该方法可以有效地消除失配并以高精度估计最佳参数，尤其是在崎areas地区。由传统的匹配策略引起；（4）采用基于最小二乘（LS）的配准的基于变换矩阵的聚类，其中消除了不匹配，并且正确对完全参与了最佳参数评估，以提高其稳定性。在模拟卫星LiDAR点云上的实验表明，该方法可以有效地消除失配并以高精度估计最佳参数，尤其是在崎rough地区。