Abstract: To overcome the limits of Digital Elevation Model in extracting erosion gully features, a method of point cloud segmentation for airborne LiDAR data based on characteristic difference is proposed. Firstly, the characteristics of point cloud of gullies are analyzed, and then the differences in surface features are obtained based on different scales after preprocessing of the point cloud. Initial gullies are segmented from point clouds with the aid of a curvature threshold. Finally, real gully point clouds are obtained based on clustering analysis. The experimental results demonstrate that gullies can be detected accurately with airborne LiDAR point clouds, and the present method provides a new idea for quantitative evaluation of gullies.

Key words: airborne LiDAR, erosion gully, point cloud, segmentation, surface feature