Automatic Extracting Road Edges from Mobile Laser Scanner Point Cloud

Information on road networks is essential in road planning and quality management. Many studies have tried to extract road surfaces, edges, and furniture from point clouds based on the advantage of rapidly collecting data from laser scanners. However, existing works have gaps in removing the road surface information for engineering purposes. This study proposed an automatic method to extract the road edge points and the road surface from mobile laser scanning (MLS). The method started to extract ground points using cell-based region growing and cell-based plane filtering. Then, contextual knowledge and Delaunay analysis were implemented to obtain the road edge points. Finally, the combination of point-based region growing, contextual knowledge, and RANSAC-based outlier removal was used to group the road edge points and remove incorrect results before generating the edge. The proposed method was tested on an MLS data set acquired from the road around Delft University of Technology, The Netherlands. Results showed that the method could extract the road edge with 95.9% in terms of the length, while the error of the road width is around 3 m when compared to the ground manually extracted from the MLS data.