The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume XXXIX-B2
25 Jul 2012
 | 25 Jul 2012


J. Lesparre and B. G. H. Gorte

Keywords: LiDAR, Point Cloud, TIN, Data Reduction, Segmentation, Reconstruction, 3D City Model

Abstract. Aerial laser scan data is widely used to produce 2 1/2D digital elevation models automatically. However, creating full 3D city models from this data is a complex task. A simplified 3D model, like a 2 1/2D model with vertical walls will be easier to produce automatically, especially when the model is not produced to look nice, but just to give realistic results in an analysis or a simulation. To be able to include vertical walls in a 2 1/2D model we extend the TIN data structure in such way that the height of a vertex is stored implicitly in a way that allows multiple heights and efficient processing at the same time. To generate the city model we first apply data reduction in planar areas of an initial TIN. Next, we segment the TIN into planar segments for the ground, walls and roof planes. Finally, we intersect the segments to obtain a 2 1/2D model with slightly slanted walls, which should be adjusted to become exactly vertical. The results for simulated point cloud data are near-perfect. Real aerial laser scan data are more challenging, but the method still gives promising results.