Automatic road extraction for airborne LiDAR data

Airborne LiDAR, as a precise and fast earth's surface three-dimensional (3D) measuring method, has been widely used in the past decades. It provides a new approach for acquiring road information. By analyzing the characteristics of LiDAR datasets as well as that of the road in the datasets, a morphological method has been proposed to automatically extract the road from airborne LiDAR datasets. Firstly, ground points are segmented from raw LiDAR data by morphological operations. The key factor in this process is how to select the window sizes in different scale spaces, and setting the elevation threshold to prevent over-segmentation in each scale space. Secondly, candidate road points are segmented from the ground points, which are obtained from previous step, by intensity constraint, local point density and region area constraint, and so on. Thirdly, morphological opening operation and closing operation were used to process the candidate road points segmented from above steps. The opening operation may effectively filter the noise areas, and greatly maintain the road detail. The closing operation may fill the small holes within the road, connecting nearby roads, and smoothing the road boundary, without signification area change. The main road can be extracted from the raw airborne LiDAR points by previous three steps. Finally, the proposed method has been verified by LiDAR data which consists of complex road networks. The result shows that the proposed method can automatically extract road from airborne LiDAR data with higher efficiency and precision.