An algorithm to automatically extract the roof points from airborne LIDAR data

Abstract

Terrain information is implied in airborne LIDAR data. Especially roof points among them are very important for 3-D building reconstruction. Therefore, this paper will present an algorithm to automatically extract roof points from airborne LIDAR data. Based on the assumption that the elevation of roof points is higher than ground elevation and the roofs are either horizontal or oblique, the roof area outlines from pure LIDAR data are roughly detected from possible roof range images, generated from LIDAR data, by image spitting and merging as well as mathematic morphology. Afterwards, some SEED regions are extracted by least squares fitting based on quad-tree spitting and merging segmentation from the LIDAR data extracted according to the detected roof area outlines. These SEED regions are used to merge adjacent LIDAR points for extracting more complete roof points by forward selection data snooping algorithm. From the experiments, the feasibility of the proposed approach will be proved.