Structured Surface Reconstruction Of Urban Buildings Based On Oblique Images

The high-precision and complete 3D reconstruction of urban architecture is a key research issue in the field of photogrammetry and remote sensing,computer vision,and computer graphics.In respect of the three-dimensional reconstruction of urban buildings,surface reconstruction based on oblique images has received much attention as the oblique photogrammetry captures information from multiple angles during the acquisition of surface information.Both the roof and facade information of building can be captured at the same time.However,there are also some problems in the three-dimensional reconstruction of urban buildings based on oblique images:Due to the obstruction and obliquity,facade information collection of some buildings is insufficient;There are some deviations in camera interior and exterior element solutions.Eventually,the reconstruction model is not regular enough and some areas are seriously deformed.It can also be noticed that there is no large-scale information missing of the building(whole roof or facade of the building)in oblique photogrammetry.With the regular structural attributes(repetition,symmetry)of the building taken into account,the reconstruction of the building can be complemented and guided using its inner attributes.Firstly,an explicit surface reconstruction based on oblique images is performed on urban scenes,then the structure of section model of the architecture is fitted and optimized using the regularity(repetition,symmetry)of the building itself,and finally the section fitting structure is used to regularize the facade and the process for complete regularized reconstruction of the building are completed.The main content can be divided into the following aspects:(1)Maximal preservation the densely matched point cloud geometry information in three-dimensional reconstruction is the basis of building regularization.This paper constructs the Delaunay tetrahedron based on the dense image matching point cloud,and a ray-constraint-based graph cut is performed on the tetrahedron,finally,explicit surface reconstruction of urban buildings is finished.(2)This paper proposes a method for structural fitting of building section.In the paper,the facades of buildings are regularized using its section structure.The complete and high-precision restoration of the regular structure of building section is the key step in the final regularized reconstruction.The method firstly orients the building model and detects the main outline structure of the point cloud of the building section,then the point cloud is divided into several parts,and finally section structure is fitted based on the segmented point cloud.(3)This paper presents a method for optimizing the section structure of a building.The fitting of cross-section regular structure will be interfered by noise inevitably.This paper proposes a method to optimize the fitting structure:the structure of the building itself has a certain regularity(symmetry,repetition).The fitting results of cross-section structure are optimized using the symmetry and repetition detected.Finally,the regularity of the fitting structure is increased and the error fitting is eliminated,and the missing feature structures in the fitting results are complemented.(4)This paper proposes a structural optimization method for building facades based on the regular structure of the building section.Based on the fitting results of the regular structure of the section,the coordinates of the vertices on the facade of the building model are adjusted,and the regularized reconstruction of the building facade is finally achieved.