The 3D Reconstruction Of City Based On The LiDAR Point-Cloud And Aerial Images

With the rapid development of GIS and the necessity of fast modeling of digital cities, it is eager to express and deal with various 3D information in the cities. In recent years, the emergence of LiDAR provides a new way to solve the problem. The data obtained by LiDAR are 3D coordinates of points on the earth surface, which express the Digital Surface Model. LiDAR has become the preferred tool to generate Digital Terrain Model and Digital Elevation Model because it can obtain high-definition and high-density 3D coordinates of earth surface automatically.According to the medium and low density LiDAR and aerial images, several aspects about urban 3D reconstruction are studied as follows:(1) With the development of LiDAR, building the TIN model with the non-uniform point-clouds of the terrain surface becomes a hot research field. Based oh the existing Delaunay triangulation method, a new algorithm of triangulation growth is presented in this dissertation. The algorithm divides the large-scale point clouds into uniform grids and determines the searching scope self-adaptively. During the process of building a Triangulated Irregular Network (TIN) model, the generated base-lines are grouped and the close-points are removed dynamically, which can improve the speed of reconstructing TIN in large-scale scenes dramatically. By searching the triangular vertices in the scope of the whole data set, the method can avoid errors caused by interpolation and the process of stitching between grids. The efficiency and effectiveness of the algorithm are verified by using real world data to build TIN model with large scale LiDAR point clouds.(2) According to the traits of urban slowly-changing terrain, based on the 3D Hough transform, restricted conditions are imposed in the dissertation, which firstly builds the basic plane of the terrain, and then extract the points on the ground in term of the position relation between the point-cloud TIN and the basic plane, and of the direction of the normal of the TIN. In order to preserve the boundary points, a new algorithm of building 2D convex hull is discussed. Firstly the points are divides into grids and the initialized convex hull is gained by seeking for the farthermost points in varied directions, then the initialized convex hull is expanded iteratively by searching the best points in local common boundary area, and finally the full convex hull from the LiDAR point-cloud can be obtained. The algorithm, which builds the final terrain TIN model by combining the points on the ground and the points of the convex hull, has low need of density of the point-cloud but has high robustness in the extraction of urban terrain.(3) The road extraction from aerial images is always a research hotspot, and the algorithm of road extraction based on dynamic programming is one of the most efficient algorithms, which is improved in this dissertation based on LiDAR point-cloud, and the robustness is also inhanced. In order to combine the aerial image and the LiDAR point-cloud data, the algorithm of matching the aerial image with the LiDAR point-cloud data is put forward. In the end, experiments are carried out to validate the algorithm.