AUTOMATIC RECONSTRUCTION OF 3D ROAD MODELS BY USING 2D ROAD MAPS AND AIRBORNE LIDAR DATA

摘要

The reconstruction of 3D road system is an important task in the geoinformatic area. The traditional topographic maps provide accurate 2D road boundaries, but are insufficient to describe the complex multilayer roads. On the other hand, the airborne LIDAR data includes the height information of road surfaces that may help to separate the road layers. Therefore, we reconstruct the 3D road models using the 2D road maps and the LIDAR data in an automatic manner. The proposed scheme comprises two stages: the establishment of 2D road networks and the reconstruction of 3D road models. First, the road boundaries in a 2D vector map are split, merged, and paired to determine the road centerlines and the widths. Those centerlines will be connected to rebuild the topology of the road network. Second, the height of each vertex belonging to the road centerlines is initially determined from LIDAR data. Then, the vertical alignment of each road segment is modified by the line fitting. Finally, the global heights of the 3D road model are fine-tuned by the least squares adjustment to maintain the continuities in height, slope, and slope difference. The test data including three types of road networks are located in Hsinchu city of north Taiwan. The scale of the road maps is 1:1,000. The mean point density of LIDAR data is 1.73 points/m2. In the validation, the 3D road models are evaluated by the manually selected reference points for the road surface. The experimental results show that the successful rate of the 2D road networking is above 97%. The modeling accuracy of the automatic reconstruction for the three road models reaches 0.10m ～ 0.20m.