Automatic building extraction from lidar data and digital maps

As the use of building models are rapidly increased for various applications, many studies have been performed to develop a practical and nearly automatic method to extract such models from various sensory and GIS data. Nevertheless, it is still a difficult problem to extract the models of large-complex buildings in particular. The purpose of this study is thus to develop a fully automatic method to extract the detail models of buildings from LIDAR data and a digital map. This extraction consisting of primitive extraction and modeling is mainly based on robust segmentation of planar patches from numerous LIDAR points. These primary primitives are used as the references to generate secondary primitives such as edges and corners and then refined based on these secondary primitives to form a complete polyhedral model. The proposed method was successfully applied to extracting large-complex buildings from real data in the test site. It can be a promising time- and cost-effective solution for a country to enhance their traditional map to include 3D models of buildings.