Evaluation Of The Potential Of Full Polarimetric SAR Data For Land-cover Classification

Land cover change is an important content in the study of global change, and land cover classification is important basis for the study of land cover change. Full polarization SAR(Synthetic Aperture Radar) data including a relatively complete physical property and structure information about the target provide a basic data for land cover change monitoring. But for now, the research work about evaluation of the ability of full polarimetric SAR data for existing land cover classification system is in its infancy, and full polarization SAR data can be independent or as a supplement of optical data used in land cover classification requires a lot of research and demonstration.Firstly, this article used the NLCD(National Land Cover Database) as a standard data to research the method of SAR data for Land Cover classification and one scene of ALOS PALSAR full polarization SAR data used as a case. First, we select a suitable classification method with classification accuracy as a guideline, and then we use different decomposition combination and band combination to classify land cover, analysis the influence of different decomposition combination and band combination on the classification results, at the same time classification results of the TM data imaging during the same period are given as a compared. The results show that a effective decomposition combination can improve the overall classification accuracy and there is a information redundancy and complementary relationship between different decomposition characteristics of SAR data, we should consider its impact on the classification results when we ues band combination to classify land cover. Band combination classification has got the highest overall classification accuracy of 71.6% and kappa coefficient of 0.6, this shows that the fully polarization SAR data can achieve good classification quality.Secondly, this article analyze the results of two scene of full polarization SAR data for land cover classification from three aspects including position accuracy, consistency of area and reliability of land cover classification. We selected the kappa coefficient, category area of correlation coefficient and structural similarity to evaluate the potential of full polarimetric SAR data for land-cover classification. The results show that the full polarization SAR data used for land cover classification have a high position accuracy and a good consistency of area but the credibility of land cover classification is not stabilization. According to different surface features, we need to choose a different band SAR data to get a good land cover classification results.