A Semi-Automatic Algorithm on Extracting Road Networks from Airborne LiDAR Data

Airborne LiDAR (Light Detection and Ranging) is a powerful remote sensing measure to derive 3D coordinates from terrain surface. Automated or semi-automated road networks extraction from laser point clouds has been a challenging work during last few decades. In this paper, combining intensity with height information of point clouds, initial road point class is discerned using few seed points and subsequently smoothed throughout filling holes technology and morphological open operations. The skeleton of road networks is acquired by virtue of image thinning method and further eliminating short road burrs with length criteria. These center points of road are tracked and connected into vector line segments. By simplifying line nodes, road network can be formed in accuracy of which completeness is 79.8% and correctness is 79%. The virtues of this algorithm enables efficiently reduce errors brought by parking lots.