Fast Building Modeling Based On Airborne LiDAR Data

3D building model is an important part of the "Digital City", and is essential to three-dimensional GIS applications. It has been widely used in urban planning, environmental simulation, cultural heritage protection, LBS, and so on. Recently, airborne LiDAR(Light Detection And Ranging) technology has been widely used in 3D city modeling. It has the characteristics of directly obtaining dense, discrete, high-precision 3D building points cloud, which is very conducive to extract the planar patches for the automated reconstruction of 3D building models. However, the accurate and fast building model automatic generation algorithm based on airborne LiDAR data is still a great challenging task in the field, and there are still some problems need to be solved, for instance, incomplete extraction of building points, over-segmentation, and complicated algorithms. Aiming at the aforementioned problems, this research developed 3D building reconstruction methods from airborne LiDAR points cloud data. The main contents are as follows:(1) A gradual advanced method was proposed to extract building roofs points from the raw LiDAR dataset. Firstly, the terrain points and non-terrain points was separated from the dataset using a morphological filtering algorithm. Then, the initial building region was derived progressively from non-terrain points by integrating the features of echo and geometrical characteristics. Finally, based on the heights and area information, building measurements was extracted. The experimental results indicate that the proposed method can extract building points effectively and the accuracy is up to 80%.(2) A progressive building roof segmentation method based on region growing algorithm was explored. In this process, initial segmentation results were firstly obtained by clustering from building points according to roof planar points cloud normal similarity. Then, non-divided points cloud were correctly assigned to the corresponding planar point set based on the distance between point and plane, and point density. Finally, complete building roofs are segmented accurately according to the method of planar patches optimization strategy.(3) A improved 3D building model reconstruction method based on topological graph was carried out by integrating building outer boundaries. Firstly, the roof ridge lines were acquired based on the topological relationship of roof planar patches.Then, the other key boundary lines, belong to two planar patches associated with ridge line, were acquired according to corresponding planar patches intersect with adjacent walls. Finally, each roof planar patchs’ closed polygon was got according to connecting rules, their combination composed the whole 3D building roof model. The experimental results show that the means and standard deviations computed by the vertical distance between point and the roof model surface is up to 0.1m and 0.08 m respectively, indicates building model fit closely with original points cloud and this method can obtain a high accuracy 3D building roof model.