The Polarimetric Decomposition And Scattering Characteristic Extraction Of Polarimetric SAR

Polarimetric SAR is an advanced synthetic aperture radar (SAR) system and is one of the important development tendencies of modern SAR systems. It can offer much more scattering information of the targets than the traditional SAR systems. This thesis focuses on the study of the fundamental theory of polarimetric SAR. The contributions of the thesis are given as follows.1) A decisive operation is proposed to solve the problem that two identical targets with different orientation angles may haveπ/2 orientation difference after deorientation. This step is added to the deorientation for totally removing the fluctuant influence of randomly distributed target orientation angles on polarimetric scattering.2) The similarity parameter between two scattering matrices can not be applied to the multi-look data of polarimetric SAR. To solve this problem, a Generalized Similarity Parameter (GSP) is proposed which is not only applicable to single-look but also multi-look polarimetric SAR data. In particular, the GSP and the similarity parameter are the same in the case of single-look polarimetric SAR. The GSP was used to modify the target detection algorithm for sea area, and the algorithm's detection performance is improved.3) The time consumption to derive the polarimetric entropy (H) and the alpha angle (α) by pixel-wise eigen-decompostion is quite tedious for very large images, andαis not stable for the case of large polarimetric entropy. To solve these problems, two parameters h and q are proposed. q has similar properties to the alpha angle, but it is stable; h also behaves similarly to the polarimetric entropy. More importantly, both the parameters can be derived very quickly. An unsupervised classification scheme based on h and q is also proposed, and its classification performance is almost the same as that by the unsupervised Hαclassification.4) Based on the properties of unitary matrix belonging to SU(3) group, a new expression for the eigenvalue decomposition of the coherency matrix is proposed. From the expression, a new compression algorithm is proposed for the multi-look polarimetric SAR data. The experiment demonstrated that the new compression algorithm has a considerably better performance on the ability to preserve the polarimetric properties of polarimetric SAR data, and that it can keep the eigenvalues of the coherency matrix non-negative after compression.5) The Freeman model-based decomposition tends over-estimate the volume scattering power, and the power contributed by surface scattering or double-bounce scattering occasionally becomes negative for some pixels. To solve both the problems, an improved three-component decomposition is proposed by employing the deorientation, a new volume scattering model, and a power constraint. By the proposed method, there is no negative power in the output the decomposition and the over-estimation for the volume scattering contribution is well reduced. The experiment and analysis demonstrated that the proposed three-component decomposition is more consistent with the actual scattering mechanisms than the Freeman decomposition.