RESEARCH ON SPACEBORNE INSAR IMAGES REGISTRATION AND REMOVAL OF THE EFFECT OF THE FLAT EARTH

摘要

Interferometric Synthetic Aperture Radar (INSAR) is a good technique to produce high resolution digital elevation model (DEM), to detect small earth motions, and so on. Therefore, it is very useful to study this technique. In this paper, some methods about data preprocessing, control points determining, image registration and removal of the phase of the flat earth are introduced, which are one big key step in INSAR technique. Usually, the original data is in a complex format. Only a few white points can be seen from this image, and it can be overcome by image histogram equalization or gray contrast enhancement. It has been tested that they improve the reliability of image registration. Image registration deals with pixel intensity transformation and geometrical transformation. In this paper, they are done separately. Histogram specialization is used to correct the former roughly, for it can satisfy requirements. The latter has two steps. At first, two images are registered coarsely by determining range and azimuth offsets because the two satellite orbital paths are very close together. Then sufficient identical points are selected by Forstner method on left image and their conjugation points are determined using normalized correlation coefficient based on the subpixel shifts between homologous areas or using the paraboloid polynomial. Finally, with the help of these control points, the geometric registration is performed by means of linear polynomials. Whose coefficients are computed with least -squares approximation. After image registration has been performed, the interferogram can be obtained by multiply each pixel in the left image by the corresponding pixel in the second image. Thus the phase difference can be gotten. This interferogram includes two parts: phase difference caused by terrain and phase difference caused by flat earth. A method called Removal of the Phase of the Flat Earth is performed. In this research, ERS-1, ERS-2 complex data in the area of Zhangbei is used. All experiments tried with methods discussed above have shown those methods are practically largely.