Research On Change Detection Algorithms Of Polarimetric SAR Image

As an all-time, all-weather microwave imaging system, Synthetic Aperture Radar (SAR)is playing an important role in civilian and military areas. Compared to single-polarizationSAR, Polarimetric Synthetic Aperture Radar (PolSAR) can describe the backscatteringproperties of targets by transmitting and receiving different polarized signals. So polarimetricSAR system is not only able to obtain a greater amount of information, but also reveal thescattering mechanisms of targets more completely and accurately, which provides datasupport and indemnification for fully exploring the information of target and improving theperformance of segmentation, classification, target detection and recognition. As an importantbranch of polarimetric SAR image processing, change detection has been widely applied todisaster assessment, search and rescue, land-used analysis and combat effectivenessevaluation, etc.Based on the study of the basic polarimetric SAR image processing theory, three changedetection algorithms are proposed, which uses different change detection features. The firsttwo algorithms regard the weighted polarization difference degree as change detection feature,but the main difference between the two algorithms is the selection of weighted coefficients,one gives all pixels the same weighted coefficient, the other one gives pixels differentweighted coefficients, which implies that joint weighted polarization difference degreealgorithm. In addition, a new change detection algorithm based on extraction of polarizationstate by selecting an optimal polarization state is proposed in the paper. In this optimalpolarization state, the scattering characteristics of same target at different phases will be moreconsistent, so we can obtain a better change detection result. Two sets of fully polarimetricSAR datas collected by U.S. UAVSAR system are adopted to test and verify the newalgorithms, and the experimental results were compared and analyzed.In different application backgrounds, interested targets are different, but man-madetargets (such as aircraft, ships, etc.) are often the focus of attention. At the end of this paper, apreliminary study on man-made targets in polarimetric SAR image is conducted. Firstly, animproved algorithm of man-made targets extraction is proposed. The improved algorithm introduced is based on the theory of polarization contrast-enhanced and the generalized Ydecomposition method, while added a second discriminating factor----the largest eigenvalueratio. Fully polarimetric SAR measured data is adopted to test and verify the new algorithm.Experimental results show that the novel method can obtain a good result with less falsedetection rate and clearly outlines. On this basis, a change detection algorithm for man-madetargets is proposed at last. In this paper we give a detailed introduction about the background,significance and process of the algorithm. The measured data is used to test and verify theeffectiveness of the novel algorithm.