Recognizing the road points and road marks from mobile LiDAR point clouds

Mobile lidar system is a cost-effective way to acquire spatial data in the urban area effectively. It can be used to generate a detailed street-level road model. As the need for Location Based Service (LBS) is increasing, the demand of understanding the city structure is growing up rapidly as well. For this reason, street-level road model is one of the most important elements to connect the geospatial objects in the urban area. The purpose of this paper is to extract the 3D road points and road marks from mobile lidar system effectively. The major works include road points selection and road marks extraction. In the road points selection, we select the lowest point as potential ground points from all points using elevation threshold. Then, we use the cubic curve fitting and point-to-curve distance to extract road points. It can remove non-ground points like cars and pedestrians. In the road marks extraction, we generate an intensity image by the interpolation of lidar intensity and create the road marking template for matching. Then, we extract location of road marks from the point clouds based on SIFT (Scale-invariant feature transform) matching. The test data acquired by Riegl VMX-250 system is located in Chiu-Chung Road in Taipei city. The accuracy of data is better than 10cm. The pixel size of intensity image is 7.5cm. The experiment results show that this method can extract ground points correctly. However, only limited road mark can be found in the preliminary result. The descriptors of the keypoints have a great effect on performance in matching.