Complex Building Edge Extraction From LIDAR Based On Undirected Plane Graph

With the continuous development of urbanization, more and more complex buildings have been rose up, which makes it particularly difficult and important to extract building edges from LIDAR data. First method takes advantage of the detection of jump points in nonparametric models since points of building edges in LIDAR data usually describe sudden local changes. The results show the efficiency of the proposed algorithm for extracting simple building edges in complex city environments. In this thesis, the algorithm for extracting edges of complex building from LIDAR data is studied. Overlooking the complex buildings, ones can see the connected planar graph which is formed by the building contours. Accordingly, the proposed algorithm is based on undirected plane graph from graph theory to model a complex building and to govern building edge extraction. Firstly, based on an undirected plane graph structure, the global and local models are established to characterize the surfaces and vertexes of buildings, respectively. From a set of starting points, the proposed algorithm tracts the data points to generate the undirected plane graph corresponding to the edges of the components. In order to determinate the property of vertices, the threshold values need to be set a priori. It is one of the important steps of algorithm implementation. The proposed algorithm is tested with raster LIDAR data and LIDAR point cloud data covering a complex building. The experimental results show the efficiency of the proposed algorithm for extracting complex building edges in city environments. Compared with the results from the Terrasolid software, the proposed algorithm is more suitable for the complex building edges extraction.