Research On The Airborne LiDAR And Visible Light Data Registration In Urban Area

There are many man-made buildings in cities, and buildings have highly random three-dimensional profiles, constitution materials and surface colours. It is difficult to use single remote sensing data to achieve accurate classification and identification of urban objects. Laser radar can obtain high-precision three-dimensional data, it is the primary approach used for the acquisition of three-dimensional urban information.Airborne Li DAR point clouds and visual image fusion are integrated in order to have three-dimensional structures, textures and colors, this will inevitably improve the timeliness and accuracy for urban object classification, target recognition and 3D reconstruction.The primary task of heterologous Data Fusion is data registration. Consequently,the paper considers real application and takes urban data as basis, takes airborne Li DAR four-dimensional point cloud data and high-resolution visible light image registration as the main contents, and carries out research from the following aspects:1. Study on the point cloud pre-processing. The discrete irregular original point cloud is transformed into elevation grid and intensity grid. For urban images, this paper proposes a road point intensity interval extraction method and an adaptive slope region growing based point cloud ground extraction method. Once iterative interpolation is performed on extraction results, the DEM and DSM data are obtained, which provides a foundation for the extraction of buildings and roads.2. In order to cater for the structural characteristics of urban roads, this paper proposes a Hough-parameter-based method for the rough registration of heterogenous images, including registration element definition, optimization of four-registration parameters and registration result verification based on homology data relevance. The method can be used to extract Hough parameters for registration under low resolution,achieving better processing results and having better applicability for aerial survey images with big data.3. This paper designs a heterogenous image registration method based on the surface characteristics of building. Point cloud data and the visible light images are adopted to extract the surface characteristics of buildings. Through sampling and weighting, it turns the surface characteristics into feature points, iterative computation of weighted minimum distance between points is carried out, in order to achieve the purpose of registration. The method can get better registration results for local contours that have identical surface characteristics.