Filtering of Artifacts and Pavement Segmentation Frommobile Lidar Data

This paper presents an automatic method for filtering and segmenting 3D point clouds acquired from mobile LIDAR systems. Our approach exploits 3D information by using range images and several morphological operators. Firstly, a detection of artifacts is carried out in order to filter point clouds. The artifact detection is based on a Top-Hat of hole filling algorithm. Secondly, ground segmentation extracts the contour between pavements and roads. The method uses a quasi-flat zone algorithm and a region adjacency graph representation. Edges are evaluated with the local height difference along the corresponding boundary. Finally, edges with a value compatible with the pavement/road difference (about 14[cm]) are selected. Preliminary results demonstrate the ability of this approach to automatically filter artifacts and segment pavements from 3D data.