An Image-Segmentation-Based Urban DTM Generation Method Using Airborne Lidar Data

Abstract

DTM generation using airborne Light detection and ranging (Lidar) data is the fundamental issue of Lidar data processing and has been massively studied. However, DTM generation is still challenging in urban areas, due to the existence of densely distributed urban features and very large buildings. Different from most point-based DTM generation algorithms, this research proposes an image-segmentation-based method for urban DTM generation. First, image segmentation is conducted using the DSM image. Next, a seed ground segment is set for each cell. Following the order of the nearest segment pair, each unclassified segment is examined by comparing the spatial correlation between the candidate segment and its nearest ground segment. This process continues until no unclassified segment remains. Based on classified ground segments, all ground points can thus be extracted and the output DTM can be obtained through postinterpolation. This method was experimented in the central Cambridge. The accuracy assessment and comparison with other Lidar-processing methods proved that the segmentation-based method produces urban DTMs with a small mean bias and limited large errors. This methodology has the potential to be applied to other areas and terrain situations. In addition to an original DTM generation method, this research works as an example that mature methods from other subjects can be employed to extend the category of Lidar-processing algorithms.