Abstract: Aiming at the problem of over-segmentation of point cloud in the process of urban road extraction caused by region growing algorithm, an improved region growing algorithm is proposed to extract ground point cloud by combining the spatial neighborhood feature information of point cloud. Firstly, data preprocessing is carried out to remove outliers far from the urban environment. Secondly, a two-dimensional spatial virtual grid is established to make rational use of the spatial locality of point cloud and reduce the scale of operation. Then, the urban road point cloud is clustered by the growth range of the mean curvature constraint region growth and the angle constraint of the fitting plane. Finally, two urban road point clouds are used for experiments, and compared with the existing region growing algorithm. The experimental results show that the proposed method can well balance the integrity and accuracy of extraction, and is practical in complex urban road point cloud extraction and urban road modeling.

Key words: vehicle-borne LiDAR point cloud, road extraction, mean curvature, tangent plane angle, region growing algorithm