Research On DSM Generation Technology Based On Video Satellite Images

In the field of remote sensing,three-dimensional reconstruction based on remote sensing image has been a long-standing problem.On this basis,the generation of highprecision digital surface model(DSM)based on remote sensing data is needed by more and more applications.In recent years,with the emergence of video satellite and the continuous progress of photogrammetry technology,DSM generation technology based on video satellite images has been widely concerned by academia.Based on Jilin No.1 video satellite data,combined with the characteristics of video satellite imaging,the complexity of remote sensing scene and the requirement of DSM accuracy,this paper carries out relevant research on DSM generation technology based on video satellite image.The main work includes the following aspects:Firstly,the characteristics of the original video satellite images are analyzed,at the same time,according to the basic requirements of the stereo matching for the video satellite images,a scheme of video satellite images preprocessing is designed.According to the basic principles of video satellite imaging and disparity statistics,it is found that the original video satellite images have the problem of the non-alignment of gazing target area in the direction perpendicular to the orbit.To solve this problem,the original video satellite images are processed by video stabilization based on feature points selected by layers,so that the video frames are basically aligned in the direction perpendicular to the orbit.In addition,image enhancement is applied to the original video satellite images in order to acquire higher image quality.Finally,processed images that meet the basic requirements of stereo matching are obtained.Then,through the analysis of traditional stereo matching algorithm,aiming at the complexity of remote sensing image texture,the traditional gray-based stereo matching algorithm can't adjust the window size adaptively,which leads to a large number of mismatched points,resulting in inaccurate disparity results.A gradient-based adaptive window size stereo matching algorithm is proposed.In addition,the causes of mismatching points are analyzed theoretically.Combined with the characteristics of video satellites,a multi-view stereo matching correction strategy is proposed to correct the initial stereo matching results.Finally,more accurate disparity results are obtained.Finally,the elevation calculation of video satellite images is studied.Firstly,from the point of view of imaging geometric model,based on empirical parameter deduction and demonstration,it is concluded that it is not feasible to use geometric model method to calculate the elevation of the target area in the image,and then to use the relevant principles of photogrammetry to do that.Specifically,firstly,the Rational Function Model(RFM)of the image is compensated and optimized based on the ground control points data,and the positioning accuracy is improved.Then,the elevation of the target area is calculated by the space forward intersection technology,and the preliminary DSM data is obtained according to the elevation results.Finally,combining with the characteristics of video frame sequence,the error of elevation calculation is analyzed,and a multi-view elevation fusion strategy is proposed to improve the quality of the video satellite image DSM.