Classification Of The Yellow River Estuary Wetland Based On Multiband And Multipolarization SAR Data

The classification of wetland research is one of the important applications of synthetic aperture radar (SAR), we need to use the information of multiband and multipolarization to make deep analysis and research. At present, the classification of wetland research in our country is more willing to use Optical Remote Sensing images while make less use of Radar Remote Sensing images. The use of SAR in our country is only aiming at single band and single polarization images without considering multiband and multipolarization information. This dissertation is supported by"908 Special Program", which choose the Yellow River estuary wetland as research area and use full-polarization SAR data of C band and L band while combining the results of"908 Special Program"and the information of Optical Remote Sensing. On the base of above work, put forward a new classification ideology in order to make better extraction of the Yellow River estuary wetland's information. Finally, according to two aspects, make comparison with the classification ability of C band and L band full-polarization SAR images. The achievements are got as follows:(1) Analyze the characteristics of C band and L band full-polarization SAR images and process the images respectively. The results prove that choose the windows of 5*5 with enhanced Lee filtering can restrain speckle noise effectively for both images.(2) Use H-A-? decomposition method to extract the characteristics of C band and L band full-polarization SAR images respectively. And on this basis, use H-?,H-?-Wishart and H-A-?-Wishart unsupervised classification methods to extract the Yellow River estuary wetland's information for both two images. The results prove that no matter which unsupervised method is easily confusing actual features for both two images.(3) Put forward a new classification ideology- the Wishart supervised classification based on unsupervised classification results, use this method to classify the different bands of full-polarization SAR images and achieve ideal results. For C band full-polarization SAR image, the overall classification accuracy reaches 92.05% while the overall classification accuracy is 67.43% for L band full-polarization SAR image.(4) According to the classification results, make comparison with the classification ability of C band and L band full-polarization SAR images from two aspects-the Freeman decomposition and Polarization Ratio. The results prove that C band full-polarization SAR image can distinguish tideland,reed wetland,plowland,river and shallow sea while L band full-polarization SAR image can distinguish plowland and reed wetland. At the end of the paper, an optimal classification scheme is raised with expectation to be solved.