3D Building Reconstruction In Large Area Urban Environment

"Cyber City" and 3DGIS are promise. They are useful in architecture, urban planning and many other potential application. With the development of "Cyber City", the demand of capturing the 3-dimension models of urban environments has been increasing steadily. Currently the 3D models of buildings is constructed mainly by hand. But to a application covered with the range of city which has thousands of buildings, it is too heavy a work to do it manually. This dissertation engages on investigating the methods of semi-automatically / automatically reconstruct the urban environments, especially the buildings. There are a number of method of methods have been presented in recent years. We compare these methods and propose a framework that can reconstruct the urban environments semi-automatically. It outputs detailed 3D building models with photo-realistic fa?ade from several photos, then positions the models on the terrain recovered from aerial photo or GIS data.The classical method of extracting 3D information from images is stereovision. It is the key point of our framework. We expound the epipolar geometry, camera calibration, and rectification and image correspondence. In a general way, the method of image correspondence can be classified into two categories: intensity-based and feature-based. As a sample of feature-based correspondence algorithm, Pollard-Mayhew-Frisby stereo Algorithm, to which central is the idea of disparity gradient, is explained here in detail.Currently, combining both spatial and temporal image information is becoming increasingly popular as a way of improving the robustness and computational efficiency of recovering the 3-D structure of a scene. Hence, the temporal stereovision emerged. We explain its thought, searching and optimization strategy.At last, we have test the PMF algorithm and temporal stereo algorithm to demonstrate the results of each.