Residential Building Reconstruction Research Based On Different Data Sources In Urban Area

Residential building modeling in an urban environment is a challenging task yet has many applications such as urban planning, simulation of disaster scenarios, cartography, wireless network planning, line-of-sight analysis, virtual tours, and so on. There are two important research foci related to building reconstruction, which are the acquisition of available data sources and improvement of modeling speed. LiDAR (Light Detection and Ranging) data and visual images are data sources available, which not only solve the problem of lacking data sources but also increase the modeling speed.This dissertation focuses on residential building reconstruction research in urban areas. Different building reconstruction methods are proposed based on different data sources, which make building reconstruction from these data sources feasible and reliable. It is also meaningful for virtual reality, urban planning, and simulation of disaster scenarios etc.The main contribution of the dissertation includes:(1) Based on the attributes of vanishing points, a building reconstruction method from terrain-level images is proposed. In this method, first, the intrinsic and extrinsic parameters of camera are calculated, and then the 3D information is retrieved. At last, the building model is reconstructed through face rendering and texture mapping.(2) Facing the sparse LiDAR data, a residential building reconstruction method is proposed, which is based on MBR (minimum bounding rectangle) and some prior knowledge of buildings. The method identifies building roof candidate points and uses MBR to isolate them. Then through roof type categorization and model matching, the buildings, such as the ones with gable roof or hip roof, are reconstructed.(3) In order to overcome several drawbacks of reconstruction based on LiDAR Data, a building reconstruction method is proposed by integrating the satellite imagery with LiDAR data. In this method, cue lines are used for building roof isolation. Complex buildings are decomposed into simple ones, which are reconstructed and rendered realistically. This was possible in spite of data noise, relatively small buildings, and very low data density.(4) In order to evaluate the building reconstruction method, different objective model quality assessment methods are defined and discussed, and then a full-reference model quality assessment method is proposed. The method uses simulated data points for reconstruction. The final reconstruction result is assessed by comparing to the facsimile of building model.