Building Model Reconstruction Of Multi-source Data Fusion

Three-dimension model reconstruction of buildings is one of the importantelements of building digital city and characterized as stereo, vivid and intuitivevisualization. Because the3DCM reconstruction of buildings needs massivethree-dimensional spatial and texture data, it is the first work to acquire, analyses anduse urban spatial information. Urban spatial data is usually acquired from aerialphotogrammetry, remote sensing, etc. As having the high precision, large range ofscanning distance and high frequency, the laser scanning is playing a crucial role inthe three-dimensional reconstruction of the building model. This thesis focuses on fordifferent characteristics of point clouds and color image, extracting useful informationof two kinds of data and integration data, for fast, accurate and automaticthree-dimensional reconstruction of buildings.In this thesis, the object of study is3D point clouds and image data acquiredfrom terrestrial scanning system. First by the conversion between scanner coordinatesystems and camera coordinate system we establish the mapping of point cloud dataand image data. Segmentation of point cloud for buildings uses a method based onreflectance image, point cloud fitting uses the least square fitting algorithm based onRANSAC. Then extracting the characteristic boundary of color data interpolatespoints. Fitting again the point cloud data on the combination of color image data, getsall plane of a building, and combined with manual handing reconstruct the buildingmodel.Study results showed that the reconstruction of buildings model using pointcloud data combination with image data is more accurate, complete, and take animplementation method quickly, accurately, and automatically. When a large numberof point cloud data is unable to get, you can get the image data manually, andintegrates the point cloud and image data for reconstruction of the building model.Experimental confirm the feasibility of the method.