Abstract:

ENVISAT ASAR Alternating Polarization images acquired at different dates，with diverse polarizations and incident angles，were used for mapping wetland vegetations of the East Dongting Lake.Based on the analysis of backscattering coefficients，an image which is made up of cross and likepolarized signals gathered from multitemporal ASAR data provides better separation between classes than a combination of likepolarized signals.Meanwhile，adopting the tone of filtered images combined with the “Contrast” texture feature based on the GLCM of unfiltered images to form the feature vector.The experimental result showed that the improved scheme has enhanced the performance of classification，overall accuracy of classification is 87.59% and Kappa index is 0.8228.Finally，the wetland located at East Dongting Lake was divided into water area，Phragmites and Miscanthus sacchariflorus，Carex and Phalaris arundinacea，woodland (Populus and Wilows)，planted land without dominant species，submerged vegetations and bare land.