Abstract: With the continuous promotion of digital city construction and the development of 3D data acquisition technology, the precision of architectural model in key areas of cities has put forward higher requirements. At present, in large-scale urban 3D modeling, the method of ground laser scanning and tilt photogrammetry combined modeling is widely used because of its high modeling efficiency, but its model precision can not meet the requirements of building modeling precision in key areas. Therefore, multi-source point cloud fusion modeling is adopted for small-scale high-precision 3D modeling. Although fusion modeling has a relatively high level of model precision, the workload of point cloud data processing in the early stage is significant, which affects modeling efficiency. This article proposes a method for fine modeling of buildings and compares it with actual data. This method not only achieves fine modeling of buildings, but also improves the automation of dense point clouds and improves modeling efficiency.

Key words: 3D laser scanning, point cloud fusion, fine building modeling