Automatic generation of E-LOD1 from LiDAR point cloud

LiDAR as a powerful system has been known in remote sensing techniques for 3D data acquisition and modeling of the earth’s surface. 3D reconstruction of buildings, as the most important component of 3D city models, using LiDAR point cloud has been considered in this study and a new data-driven method is proposed for 3D buildings modeling based on City GML standards. In particular, this paper focuses on the generation of an Enhanced Level of Details 1 (E-LOD1) of buildings containing multi-level flat-roof structures. An important primary step to reconstruct the buildings is to identify and separate building points from other points such as ground and vegetation points. For this, a multi-agent strategy is proposed for simultaneous extraction of buildings and segmentation of roof points from LiDAR point cloud. Next, using a new method named “Grid Erosion” the edge points of roof segments are detected. Then, a RANSACbased technique is employed for approximation of lines. Finally, by modeling of the rooves and walls, the 3D buildings model is reconstructed. The proposed method has been applied on the LiDAR data over the Vaihingen city, Germany. The results of both visual and quantitative assessments indicate that the proposed method could successfully extract the buildings from LiDAR data and generate the building models. The main advantage of this method is the capability of segmentation and reconstruction of the flat buildings containing parallel roof structures even with very small height differences (e.g. 50 cm). In model reconstruction step, the dominant errors are close to 30 cm that are calculated in horizontal distance.