Study On Change Detection Technology Based On Polarimetric SAR Data

Change detection is to extract change information by comparing and analyzing multi-temporal data. The technology can acquire real time change information. So it has been widely used to monitor agriculture,forest,land use,urban management planning,natural disaster etc. Polarimetric SAR not only has all-day long and all-weather imaging work ability, but also can obtains abundant intensity and phase information. They are the advantages of polarimetric SAR. Study on polarimetric SAR change detection technology is significative for getting change information quickly. Therefore, the dissertation mainly focuses on developing algorithms of research change detection measures from polarimetric SAR data and is organized as following:Firstly, the theoretical foundation of polarimetric SAR is introduced, which includes the formulation of polarization states, the characterization of polarimetric backscattering and polarization synthesis. This part aims to provide some preparing knowledge for change detection.Secondly, author analyses the distuibution of distribution of covariance or coherency matrices, make use of statistical hypothesis test to build likelihood ratio parameter and describe the change degree of features.Thirdly,author enhance the change phenomenon through optimal polarization and OPCE, and then build diff image by logarithm ratio measure. After processing as this, the change areas are integrated and the speckle is reduced effectively.Fourthly, there are two measures to extract change information. One is GGM-EM.it estimate the parameters of generalized gauss mixture distribution model by expectation maximization method, then extract change information by the error rate minimum bayesian theory. The other is probabilistic relaxation iteration algorithm. It is initialized by the result of GGM-EM, then considering the relationship between 8 pixels and optimizing the result of GGM-EM. The speckle is reduced and the accuracy is increased.Last, author compared the results with threshold method, and proved the methods are effective.