Abstract:

The construction of surface model by using dense point cloud obtained from oblique image is one of the core of 3D reconstruction based on oblique images. For the problems of low modeling efficiency and unreal surface selection in the current dense point cloud surface model reconstruction, a new method of urban dense point cloud surface model reconstruction based on graph-cuts algorithm is proposed. This method is used to reconstruct the surface model of the urban dense point cloud. First, the aerial triangulation of the unmanned aerial vehicle (UAV) image is done by the preprocessing software, and the dense point cloud is generated by the results of the aerial triangulation, then the corresponding edges are added to the dense point cloud, and the 3D point clouds are selected and merged according to the distance. Finally, a graph-cuts problem is established based on the tetrahedron and triangulation formed by the 3D point cloud, and an optimal dense point cloud surface model is obtained by solving the graph-cuts problem. In order to prove the feasibility and effectiveness of this method, the UAV image data in urban areas is used to reconstruct the urban dense point cloud surface model, and the experimental results show that the method has the advantages of good feasibility, good modeling effect, fast processing speed and so on.

Key words: urban 3D, oblique images, graph-cuts theory, dense point cloud, surface model reconstruction