Research On Stereo Matching And Generating DSM For ZY-3 Satellite Three-line-array Imagery

Nowadays,with the technology of remote sensing satellites for earth observation changing with each passing day,the imaging method of remote sensing image is not only confined to the frame image,the linear array CCD sensor imaging method has become a hot spot and difficult point in domestic and foreign scholars.As China's first high resolution civilian stereo mapping satellite,ZY-3 satellite,the sensor payload is the linear CCD camera.This paper based on the linear array push scan camera imaging modalities,with ZY-3 satellite three line array CCD image data as an example,according to the imaging principle,through matches the characteristics of the different scaled nadir and backward images,establishing rational function model,in the absence of control point data and DEM,with the RPC parameters of the satellite image from ZY-3 satellite for three-dimensional coordinate calculation of ground,so as to generate the digital surface model.Because of the ZY-3 surveying and mapping satellite in the process of flight,by the influence of various factors,will inevitably lead to the existence of a certain image error,before the image is analyzed,the remote sensing image should be corrected and enhanced.This paper use the method of adaptive smooth filtering and Wallis filtering for image enhancement.According to the particularity of the satellite three line array CCD imaging,proposed the multi-scale space Harris and SIFT combination of improved image detection method,which uses Harris corner detection algorithm in the generation process instead of the SIFT algorithm to generate the extreme points of the process,and to ensure that the feature points with scale invariance.Firstly,nadir and backward images of ZY-3 satellite were used to construct the Gauss scale space,then extracted feature points in Gauss scale space of nadir and backward images with Harris scale invariant detector.Due to some inevitable operation of pseudo feature points in multi-scale space,non feature points should be removed by excluding 8 neighborhood pixels similar method are improved by Harris method,then describe the extracted correct feature points with feature descriptors in SIFT algorithm.By Euclidean distance to coarse match feature points of the foresight and back sight image,then use the least squares matching to eliminate matching points of gross error.Under the different resolutions of ZY-3 nadirward and backward images,through this method,can improve the robustness of feature points in the imagery matching,to ensure the accuracy of image matching,and get some effect in the match time and computational complexity.Next,a rational function equation is established by using the two dimensional coordinates of the homonymy points of the image matching and the RPC parameter file.By means of the method of the positive rational function model to calculate the three-dimensional coordinates of the ground point corresponding to the image point then generate DSM.The experiment was carried out by using ZY-3 satellite in Beijing area of three line array image data,verified the validity and applicability of the algorithm in this paper.