Complex Digital Building Model Generation through the Integration of Photogrammetric and Lidar Data

This paper is concerned with the integration of photogrammetric and LiDAR data for accurate generation of complex Digital Building Models (DBM). Some research has been conducted utilizing either high overlap/sidelap photogrammetric data or high point density LiDAR data in order to specifically cater to DBM generation. In contrast, the proposed methodology is designed to utilize imagery and LiDAR data acquired in conventional mapping missions for building detection and reconstruction. The building detection is established using LiDAR data alone through several intermediate steps: ground/non-ground separation, building hypothesis generation, and building primitive generation. The planar patches constituting the building rooftops are used as building primitives and are derived from a LiDAR segmentation procedure. Due to the irregular nature of the LiDAR data, the planar patches have irregular boundaries. Therefore, a boundary refinement procedure is developed through the integration of imagery and LiDAR data. Using straight lines derived from the involved imagery, 3D lines are established from a matching process on the planar patches defined by the LiDAR data. Afterwards, these 3D lines are considered as candidate lines for building boundaries and divided into non-overlapping groups by checking their geometric similarities. A spectral similarity measure is established to define the most appropriate building boundary among the matched lines for each of the involved groups. Finally, the process concludes by a manual editing procedure to add any missing lines or to remove incorrect lines. The performance of the proposed system is investigated through experimental results involving real datasets.