基于平面几何特征约束的城市机载LiDAR点云精确配准算法

An Accurate Registration Algorithm for Airborne LiDAR Point Clouds in Urban Scenes Based on the Plane Geometric Feature Constraint

针对目前城市机载LiDAR点云数据在配准过程中存在的自动化程度低、效率低、配准结果易陷入局部最优等问题,提出一种基于平面几何特征约束的点云配准算法。该算法首先基于渐进形态学滤波算法提取非地面点,并依据点云平面度与粗糙度特征提取建筑物点云,然后通过欧式聚类与RANSAC算法提取建筑物点云平面特征,接着通过计算重心KD树与Hausdorff距离构建平面特征的相似性系数矩阵,同时引入双向一致性约束规则与平面法向量余弦相似性测度进行平面特征匹配,最后依据平面几何特征约束条件构建点云四元数坐标转换模型,实现了点云数据的精确配准。实验结果表明,相对于传统的人工选取同名平面特征的方法,本文算法的匹配效率、准确率和配准精度均有显著提高。

Aiming at the current problems of poor automation,low efficiency and easy convergence to local optimal solution in the registration process for urban airborne LiDAR point clouds,a point cloud registration algorithm based on the plane geometric feature constraint is proposed in this paper.Firstly,the non-ground points are extracted by progressive morphological filtering algorithm,and the building points are separated from the non-ground point according to the planarity and roughness of the point clouds,and then the building roof planes are segmented through the European clustering and RANSAC algorithm.Secondly,the similarity coefficient matrix of planar features is constructed by calculating the KD tree and Hausdorff distance,and the bidirectional consistency constraint rule is introduced to match the plane feathers with the cosine similarity measurement of the normal vectors.Finally,the quaternion coordinate transformation model of point clouds is established according to the constraint of planar geometric features.The experimental results show that the matching efficiency,accuracy and registration precision of the proposed algorithm are significantly improved compared with those of the traditional method of manually selecting the correspondence planes.