车载激光扫描的三维道路自动提取方法

车载激光扫描系统获取的复杂道路环境点云数据量大、目标复杂，难以有效提取出道路的点云。本文通过分析扫描线上激光点云的空间分布和统计特征，提出一种适用于复杂道路环境的道路点云自动提取方法。该方法首先根据点的扫描角度或GPS时间信息提取扫描线；利用移动窗口法进行高程滤波，提取地面点云，然后采用基于路坎模型的移动窗口法提取路坎点；利用局部区域相邻扫描线的相似性特点，对提取的路坎点云进行跟踪和优化；最后利用优化后的路坎作为道路的边界实现道路路面精确提取。经过实验和分析，该方法不仅适应于有固定道路宽度的结构化道路提取，同样适用于无固定宽度的复杂道路提取。

Automated Extracting 3D Roads from Mobile Laser Scanning Point Clouds

This paper proposes an automated method for extracting road from Mobile Laser Scanning point clouds. Firstly, the scanning lines were extracted from laser scanning point cloud according to GPS time of point. Then, a moving window operator was adopted to filter non-terrain points. Secondly, curb points were detected from terrain points based on curb models. Finally, an algorithm was implemented to track and refine curb points with both global consistency and local similarity properties of road. Experiments were undertaken to evaluate the validities of the proposed method, which proves a promising solution for extracting roads from mobile laser scanning point clouds.