AUTOMATIC BUILDING EXTRACTION FROM LIDAR POINT CLOUD DATA IN THE FUSION OF ORTHOIMAGE
Keywords: Building, Segmentation, Mean Shift, Image, LiDAR, Point Cloud
Abstract. Three-dimensional building models are important in various applications such as disaster management and urban planning. In this paper, a method based on the fusion of LiDAR point cloud and aerial image data sources has been proposed. The first step of the proposed method is to separate ground and non-ground (that contain 3d objects like buildings, trees, …) points using cloth simulation filtering and then normalize the non-ground points. This research experiment applied a 0.1 threshold for the z component of the normal vector to remove wall points, and 2-meter height threshold to remove off-terrain objects lower than the minimum building height. It is possible to discriminate vegetation and building based on spectral information from orthoimage. After elimination of vegetation points, the mean shift algorithm applied on remaining points to detect buildings. This method provides good performance in dense urban areas with complex ground covering such as trees, shrubs, short walls, and vehicles.