The Processing And Application Of UAV Image Matching Point Cloud

As a powerful supplement to traditional photogrammetry,unmanned aerial vehicle(UAV)has played an indispensable role in national land survey,rural land-use rights verify,smart city construction and emergency rescue in recent years.UAV image matching point cloud is a 3D feature point set obtained by dense matching,but its quality is poor due to the weak quality of the image,the distortion of the lens,the limitation of matching algorithm and the influence of various errors.For example,the boundary information of the point cloud is fuzzy and there exists large amounts of rough and redundant points,all of these will greatly affect the quality and application of DSM and DOM.Based on overall analysis of the characteristics of UAV image matching point cloud,this paper deeply studies data organization,simplification,filtering,noise rejection of the matching point cloud and the generation of true ortho-image.The main work and innovations are as follows:1.This paper summarizes the current situation of the image matching point cloud from data organization,thinning,filtering and analysis the characteristics of UAV image matching point cloud.2.A data organization model for UAV image matching point cloud based on virtual grid and octree is proposed,and the point cloud data search and scheduling are achieved by combing internal with external dynamic scheduling strategy.Experimental results show that our method can perform the organization and dynamic scheduling of one billion point cloud under the condition of limited memory resources.3.Aiming at solving the problem of data redundancy of dense image matching point cloud,a simplified algorithm based on Hilbert fractal curve is designed.Experimental results show that the proposed algorithm is effective,which reduces point cloud redundancy and improves processing efficiency.4.The method of gross error elimination based on statistics and window height difference is used to solve the problem that the matching point cloud contains error point.The moving surface fitting filtering scheme with robust estimation is designed to improve original algorithm and eliminate the artificial objects in point cloud.5.The boundary of the UAV image matching point cloud is always fuzzy and adhesion,resulting in a poor quality of DOM.To solve this problem,DLG is used to optimize the building point cloud,and then the true ortho-image is generated by the optimized point cloud.