Polarimetric SAR Image Classification Based On K-wishart Distribution

Polarimetric Synthetic Aperture Radar (POLSAR) is a muti-channel imagingradar system.Compared with the single polarized SAR, POLSAR includes moreinformation of targets and polarimetric, which can provide more surface featureinformation and classification in polarmetric SAR image classification. so the studyof POLSAR image classification plays an important role in POLSAR dataapplication. This paper mainly study SAR image clasification methods based on thecombination of statistics characteristics and physical scattering characteristics.Combining with the cluster analysis methods and the polarization scatteringmechanism, this paper proposed several improved classification methods based onK-wishart distribution, which has been shown to be good fits to real POLSAR data:1. We proposed a method based on the Freeman-durden decomposition anddata distribution characteristics. Apply the three scattering powers to achieve theinitial classification based on the main scattering power at first. Then use the datadistribution characteristics to classify the image into more classes. F inally, the imageare classified by an iterative algorithm based on a complex wishart density functionto improve the classification accuracy. Compared the classification results with othertraditional decomposition methods,the proposed algorithm idea can obtain betterresults and it is easy to understand and implement.2. Based on the work of the former chapter,we impoved the the traditional andclassical algorithm H/α-wishart and constucted the K-wishart classifier.Thenproposed a method based on Cloude decomposition conbining with thisclassifier.Compared with the Wishart distribution,the K-wishart distribution canbetter descripte the POLSAR data.The classifier based on this distribution has amore favorable performance.The experimental results shows that on the introductionof the K-wishart distrbution,the classification result is further improved.3.We proposed a method based on the MRF and K-wishart distribution. Thealgprithm used MRF model to calculate the a priori probability based on K-wishartclassifier, fixed the K-wishart distance based on the priori probability,and added thesample selection strategy in the iteration process of K-Wishart,Selected a prioriprobability sample to participate in the calculation of the cluster center, and madethe parameter estimation and the selection of clusteringcenter more accurately. Theproposed method using the MRF further considers the spatial information on thebasis of the K-wishart distribution.The experimental verified that the consistency of terrain classification area consistency and correct rate are both increase.