| As one of the 3S technologies,photogrammetry and remote sensing technology is an important supporting technology to obtain the geometric information of the surface space target,reconstruct the three-dimensional model of the surface and construct the digital earth.For the satellite line array images with large image amplitude and high resolution,the realization of image intensive matching is an important step to extract the three-dimensional information of the surface space.The key is to reduce the search range of the same name point from two-dimensional to one-dimensional through the epipolar line constraint,so as to ensure the efficiency and reliability of the matching process.In this paper,the epipolar line model of line array images is studied,and a linear epipolar line model and its parameter estimation method based on the principle of minimum vertical parallax are proposed,which can be used to generate kernel images and 3D reconstruction of various satellite linear array images at home and abroad.The main work and innovation of this paper are as follows:(1)Based on minimum vertical parallax,to optimize the extended epipolar line model,a new image linearized epipolar line model and its determination method are proposed.For single center projection stereo pairs,the results of the new geometry are in good agreement with the traditional epipolar line model;Experiments on ZY3,PLEIADES,SPOT-5,Quickbird and other satellite linea array images show that the vertical parallax epipolar lines obtained by the new model in different image ranges is significantly smaller than that obtained by the conventional extended epipolar line model,and the accuracy of the new model improves from 31% to 72% in the range of 5000×5000 pixels.In addition,due to the differences in stereo imaging methods,sensor geometric characteristics and specific imaging processes of different satellites,the variations of the epipolar line inclination of each image show different laws.ZY3 images can be regarded as approximately parallel,while SPOT-5 and QuickBird images show obvious variations of the epipolar line inclination.(2)According to the variation characteristics of the dip Angle of the epipolar line,the adjustment estimation models with seven parameters and eight parameters are proposed.A certain number of joint points within the scope of the image block were obtained through SIFT feature matching,and error equations were constructed by using the joint points,and model parameters were solved by adjustment.The experimental results of simulation data and actual satellite image data show that the error accuracy of the eight-parameter kernel line model is improved to some extent compared with that of the seven-parameter kernel line model,and the improvement effect of the eight-parameter epipolar line model is especially obvious for the images with great changes in the tilt Angle of the epipolar line,such as SPOT-5 and QuickBird.A simplified seven-parameter model can be used for images with epipolar line nearly parallel to ZY3 and PLEIADES.(3)Based on the adjustment results of the epipolar line model,the epipolar line image was generated from the original images of ZY3,SPOT-5 and QuickBird.The high-density matching was performed by SGM method and the surface dense point cloud was generated to construct the 3D surface model.The research results of this paper optimize the 3D modeling technology of satellite linear array image to a certain extent,and it is expected to be further applied to the new generation of domestic surveying and mapping satellites for 3D earth reconstruction... |
PDF Full Text Request