Research On Real Shot - Like Image Generation Method Based On Airborne LiDAR Point Cloud

Digital orthophoto map (DOM) is a very common part of spatial datasets with rich texture and geometric map attributes. A normal orthophoto is made on basis of digital elevation model (DEM) which doesn’t include the buildings, vegetation etc. This results in an image where buildings are leaning away from the image center, and the relief displacement of the buildings can be so large, that will obscure the terrain and objectives next to them. Especially in built-up urban areas, the buildings’tilt and occlusion phenomena will be more serious, so the DOM lose geographic reference value. True digital orthophoto map (TDOM) is a more advanced ortho products proposed under the above background. TDOM is correct on basis of digital surface model (DSM) which not only contains the terrain, also contains the artificial features such as buildings. So in the TDOM, the building can be correct to the right position, then the area blocked by buildings can be recruited through occlusion detection and texture repair. The traditional photogrammetric method to generate DSM not only has a long production cycle, high cost and low precision, is not enough for true ortho-rectification and occlusion detection. However the airborne laser radar (LiDAR) system can obtain high precision 3D laser point cloud data of the survey area, and provides high-quality data sources for DSM generating.The key to generate TDOM is high quality DSM production and occlusion detection. The purpose of this study is generating TDOM from aerial photographs and DSM. The main reseach content includes the following aspect:(1) Generating DEM from LIDAR point cloud.DSM can be regarded as the combination of DEM and digital building model (DBM). For the DEM production, a progressive TIN encryption filtering algorithm has been proposed for the discontinuous ground point cloud extraction in this paper, then generate DEM by Delaunay triangulation of ground point cloud.(2) Generating DBM from LIDAR point cloud.The DBM generation from LIDAR point cloud is divided into two steps:the first is to separate building point cloud from the original LIDAR point cloud; the second is to generate DBM from building point cloud. For the building point cloud extraction, this paper proposed a roof point cloud extraction algorithm based on 3d hough transform. For the DBM generation, the method to generate DBM based on the Invariant moment of point cloud has been studied in this thesis, thus avoiding the jaggies at the edge of DBM caused by the point cloud’s blindness and unevenness.(3) Occlusion detection based on DSMOcclusion detection is an important part for TDOM generation. Through discussion and analysis of existing occlusion detection method, this paper proposed a constraints region z-buffer algorithm to detect the shaded area. This algorithm just work at the building relevant area, thus saved much unnecessary calculation and improved the detection efficiency.(4) Texture repairThis paper discusses the filling texture selection’ selection method based on the minimum imaging angle in detail. This method uses the texture on adjacent image whose imaging angle to shaded area is smallest to repair shaded area. The smaller the imaging angle is, the better the repair effect is.The method proposed in this paper to generate TMOD with LIDAR point cloud and images achieved the true ortho-rectification of a single width rule building image. However, the styles of building vary widely in most cities, how to improve the universality of TDOM generation method still requires lots of depth research.