LiDAR Point Cloud Assisted Aerotriangulation Of Urban Airborne Image

Informatization and interlligentialize is the basic characterize of photogrammetry in the information era,the convenience of data procurement and the effectiveness of data process enhance the sensing capability of geo-information.The increasing requirement of high accuracy urban 3D geo-information is significant in the process of urbanization.As the primary technique for 3D geo-information collecting,photogrammetry depends heavily on the high accuracy artificial measured control point,the acquisition of such point needs massive manpower,physical resource and which is extremely time consuming.Since the photogrammetry tecnique is a multi-stage process,the production cycle can be delayed and thus reduce the efficiency.For such reasons,the control information automatic acquisition technology attracted increasing attentions in photogrammetry field.The artificial control points are usually used to airborneimagery orientation in the aerotriangulation process.Although the Position and Orientation System(POS)is widely used in airborne camera,but lots of research have been proved that the precision of low-cost POS is far from reaching directly use for high precision applications.In fact,the POS measurement is often used as auxiliary data in the aerotriangulation,artificial control points is essential in realisitic production.Control information can be extracted from existing high precision geo-data and then used for newly acquired imagery position and orientation,this approach is widely explored by researchers.The existing geo-data conclude imagery with accurate exterior orientation parameters,orthoimages,LiDAR point cloud,digital linear graph and the LiDAR point cloud is the most widely used.LiDAR points are appropriate for imagery orientation because which is dense enough with high positioning accuracy,airborne imagery can be registered to LiDAR data to enhance positioning accuracy.LiDAR points can also use to making up the shortcomings of image based dense match point cloud.Therefore,LiDAR point cloud can improve both of the quality and efficiency of photogrammetry products.In summary,the study of LiDAR points controlled airborne imagery triangulation is meaningful,on one hand the LiDAR points can improve the image positioning accuracy,while on the other hand the LiDAR points can be mixed use with photogrammetry products for extensive applications.This research focuse on LiDAR point cloud controlled urban airborne imagery triangulation.For various airborne imagery,a robust data process framework is developed to realize automatic airborne imagery orientation.Under this framework,data analysis has been done to evaluate feasibility of LiDAR data as control information,then the method for LiDAR points controlled free-net construction and airborne imagery triangulation is studied,we conduct extensive experiment on realistic data sets to prove the effectiveness of the proposed method.In particular,the main contents of this research are as follows:(1)Precision evaluate of LiDAR point cloud is studied,the feasibility of LiDAR points as control data should be strictly discussed since the control data is supposed to have very high positioning precision.(2)Efficiency bottleneck of classical Structure from Motion(SfM)workflow is analyzed in detail.To solve these bottleneck problems,LiDAR data is introduced to the SfM workflow as auxiliary data to enhance the data process efficiency.Main research content of this part concludes image adjacent matrix evaluate,spatial verify and incremental free-net construction.(3)3D line reconstruction method based on multi-view geometry is studied.For robust registration of airborne imagery and LiDAR data,line feature is fully considered in different procedure in the proposed framework.This dissertation present a 3D line reconstruction method based on multi-view geometry and LiDAR data assistant.The image-based 3D line and LiDAR feature line are registered by iterative cloeast line(ICL)method.(4)LiDAR point cloud controlled free-net absolute orientation is studied.This dissertation divided the absolute orientation as two steps,the first step is the rigid registration between free-net and LiDAR point cloud by using ICP and ICL combined method,the second step is the non-rigid registration by using point-point and point-line constrainted bundle adjustment,in which the point-point constraint represents the distance between tie point in free-net and LiDAR point cloud,the point-line constraint represents the distance between back-projective LiDAR feature line and corresponding image feature line,respectively.