Research On Three-dimensinoal Reconstruction Technology Based On Satellite Remote Sensing Images

With the excellent data acquisition and transmission capability of high resolution remote sensing satellite,the technology of high resolution remote sensing satellite plays a more and more important role in the field of national economy construction,and the three-dimensional reconstruction technology based on high resolution remote sensing satellite has been closely concerned by all countries all over the world.The purpose of this paper is to establish a process to complete three-dimensional reconstruction based on high-resolution satellite remote sensing images.It mainly focuses on the stereo matching of the binocular vision system,the three-dimensional reconstruction based on the rigorous sensor model,and the three-dimensional reconstruction based on the rational function model.Firstly,complete the stereo matching of a pair of high-resolution satellite remote sensing images,then obtain dense disparity map.Stereo matching is the core technology in three-dimensional reconstruction,and the accuracy of the reconstruction results directly depends on the matching results.The stereo matching algorithm used in this paper is an improved semi-global stereo matching algorithm.As we all know,the semi-global matching algorithm perform bad in weak texture area.In order to overcome the shortcomings,this paper deeply studies the relationship between the tree-weighted belief propagation algorithm and the semi-global matching algorithm,and adopts a more global traversal order to calculate the aggregate matching cost.By analyzing the experiment results,the stereo matching algorithm used in this paper can effectively eliminate the atreaking artifacts of the semi-global and obtain a more accurate disparity map.Secondly,complete the three-dimensional reconstruction of the WorldView-1 high-resolution satellite remote sensing image based on the rigorous sensor model.This paper detailed introduces the imaging geometry principle of WorldView-1 satellite then discusses the coordinate systems involved in the construction of the rigorous sensor model and the transform relationships between of them.On this basis of this,the rigorous sensor model of WorldView-1 satellite is established,and verifies its correctness and effectiveness through real ground control points.In order to improve the positioning accuracy of the rigorous sensor model,this paper carefully analyzes the error types and its propagation rules in the satellite imaging process,and uses the bundle adjustment method to modify the rigorous sensor model.Based on the disparity map of satellite images,this paper uses the forward intersection method to realize the high precision three-dimensional reconstruction based on the rigorous sensor model.Finally,complete the three-dimensional reconstruction of the WorldView-1 high-resolution satellite remote sensing image based on the rational function model.Detailed description of the positioning principle of the rational function model.Since the rational function model is the reparameterization of the rigorous sensor model,there is a large systematic error in the rational function model.This paper adoptes the compensation method based on affine transformation in image space to modify the rational function model in order to improve the positioning accuracy.Considering the difficulty of acquiring ground control points,this paper proposes a method for correcting rational function model based on SIFT matching points,and verifies the feasibility of this method through experiments.In summary,the three-dimensional reconstruction method scheme of the high-resolution satellite remote sensing images proposed in this paper can obtain high positioning accuracy on the WorldView-1 satellite remote sensing images,and the minimum average elevation error can reach 2.251 meters.