3D Reconstruction Technology of Urban Buildings Based on Computer Vision and Point Cloud Data

Under the background of rapid development of geographic information system (GIS) technology, how to effectively describe and process all kinds of three-dimensional information in urban space has become an urgent problem to be solved. This paper focuses on the construction of urban 3D modelling system. The point cloud data obtained by laser scanning or photogrammetry can provide accurate spatial information of urban buildings, including the shape, size, location and structure of buildings, etc. Using these point cloud data, the key technologies such as the fully automated extraction of digital elevation model, the representation of architectural model, and the reconstruction of urban 3D geomorphology are investigated. A large number of experiments are carried out to verify the proposed theory and method. The experimental results show that the average deviation of the models for window 1, billboard, door, floor corner and window 2 is 6.2mm, which is much lower than that of BIM (building information modelling) technology and computer vision and point cloud data.