As an active imaging sensor, Synthetic Aperture Radar (SAR) was born in the 1950s., Compared with optical sensor, it can acquire high resolution radar images all the day and all the time. After doing interferometric processing with the SAR SLC images of the same area acquired at different time, the results can be used to generate digital elevation model (DEM) and to detect the subtle deformation of the earth surface. Along with its advancement, SAR technology has been largely applied in many areas such as topographic mapping, hazards evaluation, geology, geodesy, land environment, atmosphere sciences and so on. This dissertation is mainly concentrated on such key issues in interferometric processing as: How to generate high quality interferograms efficiently and robustly, fast and accurate estimation of SAR images coherence, interferogram filtering, and the combination of ERS and Envisat images for cross interferometry. The main results can be summarized as follows: 1) A fully automatic interferogram gene:ration from coarse registration to fine registration is obtained. In order to calculate the correlation efficiently, a new fast correlation calculation method is proposed. It yields satisfying results even for the complex correlation calculation. Some particular efforts are dedicated to improving the registratiol: accuracy in the fine stage, Besides, a correlation ratio based pyramid method is used to select the matching points by verifying the consistency of offsets. In particular for the polynomial interpolation kernel, the Homer principle is adopted to save a large quantity of computations. 2) The accuracy of SAR images co-registration based on the coherent and non-coherent correlation is investigated. A new method is proposed to reduce the decorrelation effect where the phase of the flat earth is calculated first, then the flattend SAR complex irrtages are used for registration. The experiments validate the proposed method. 3) The flattening effects using the flat earth model and the elliptic model are compared. It is shown that the polynomial fitting of the flat earth phase can substantially improve the computational efficiency. Coherence magnitude estimate based on the sample coherence magnitude, the flattened sample complex coherence and the average of the flattened sample complex coherence are studied for their impacts on the interferometric SAR coherence estimation. A fast coherence estimation method is proposed. Its computation load is almost independent of correlation window size. 4) The problem of interferogram filtering based on Wavelet Transform is studied. The wavelet filtering method based on hidden Markov tree model (HMT), which has been successfully applied to digital image filtering, is
修改评论