Abstract: Vehicular light detection and ranging (LiDAR) technique has great application prospects in smart roads because it can quickly and efficiently obtain the geographic information of the road and both sides. Massive point clouds collected by the LiDAR technique achieve ‘what you see is what you get’. However, due to its large amount of data and many noise points, the relevant methods of road centerline extraction are not mature enough. A point cloud data processing method for centerline extraction of high-speed overpass was proposed, in which multiple filtering algorithms (gradient filtering, Gaussian filtering and bilateral filtering) were used to filter the off-road point clouds, the Alpha shapes algorithm was used to identify the road boundary and the B-spline curve fitting algorithm was finally introduced to fit the road centerline. The effectiveness and feasibility of the proposed method were verified by the vehicular LiDAR data of a high-speed overpass in a case study. Relevant research results can be used in the fields of smart road reconstruction, expansion and maintenance.

Key words: highway engineering, smart road, road centerline, vehicular LiDAR, point clouds, highway overpass, B-spline curve

摘要： 车载激光扫描（light detection and ranging，LiDAR）技术因其可快速高效获取道路及两侧的地理信息，在智慧道路中有很大的应用前景，其所采集的LiDAR点云，实现了“所见即所得”，但其数据量大和噪点多等特点导致道路中线提取相关方法还不够成熟。提出了一种面向高速立交桥中线提取的点云数据处理方法：采用多重滤波算法（梯度滤波、高斯滤波和双边滤波）过滤非路面点云，应用Alpha shapes算法识别道路边界，最终，引入B样条曲线拟合算法拟合道路中线。通过实例高速立交桥车载LiDAR数据验证了所提出方法的有效性和可行性。相关研究成果可服务于智慧道路改扩建和维养等领域。

关键词: 道路工程, 智慧道路, 道路中线, 车载LiDAR, 点云, 高速立交桥, B样条曲线