Study On The Key Techniques Of Polarimetric Synthetic Aperture Radar Interferometry

Polarimetric interferometric synthetic aperture radar (PolInSAR) is an advanced SAR system. Since polarizations of electromagnetic waves are employed, polarimetric SAR interferometry has more extensive applications than traditional SAR interferometry. Although some researchers have done much work since the first paper on polarimetric SAR interferometry was published 11 years ago, we still need to solve some problems in theory. For developing the applications of a PolInSAR, this thesis focuses on some key techniques, including image registration, data denoising, phase unwrapping, and tree height estimation. The following innovations are obtained in this thesis.1) An image registration method for PolInSAR images based on the similarity parameter is proposed: a new registration measurement is defined by the correlation coefficient between the coherent polarimetric scattering vectors, and the method performance is analyzed theoretically. It is proved to have better robustness and accuracy than traditional methods using only interferometric information. The noise caused by mis-registration is reduced and it is beneficial to the later interferogram extraction.2) A denoising method for InSAR data based on the amplitude- optimization (AO) principle using polarimetric information is proposed. By searching the optimal polarimetric channel to maximize the amplitude of the InSAR signals, the phase between them is improved. The mathematical model is built and solved analytically. A physical explanation is supported by the similarity parameter. As an extension, an intensity-optimization (IO) based method is proposed for multi-look input data. The performance of the proposed method is better than those of the coherence-optimization based methods, especially in weak signal areas.3) Two filters, the fuzzy filter and the AO based filter, are proposed for InSAR data filtering. Both of them make use of the relationship between the amplitude and phase of the complex signal. In the fuzzy filter, neighborhood information is used to improve weak signals by interpolation, and then the phase is processed by weighted median filter. In the AO based filter, the AO model is introduced to obtain the optimal filter coefficients. They are proved to have better denoising ability than traditional methods. Particularly, the latter filter uses the neighborhood information with the highest efficiency.4) Two phase-changing techniques, based on the phase segmentation and the referenced surface respectively, are proposed to overcome the difficulty caused by steep topography. The former regards the phase unwrapping as an image segmentation problem, segmenting and combining the phase reasonably. The latter regards it as"topography effect removal"problem, constructing a referenced surface to simplify phase unwrapping. Experiments demonstrate that combining the two techniques can improve the effectiveness and accuracy for complicated topography inversion.5) In the tree height inversion problem, a volume coherence estimation method is proposed. By considering the argument range of the coherence, an approximate analytical solution of the best volume coherence estimation is obtained. Experimental results show that this method can be used to improve the accuracy of the tree height inversion, demonstrating its advantage over conventional estimation methods.