Land-cover classification using SAR and MS image with ANN classifiers

ABSTARCT: In this work, gray-level co-occurrence matrices (GLCM) have been used to quantitatively evaluate statistical textural parameters for a SAR image and to generate a filtered image to feed to ANNs for classification for land cover. Prior to performing the textural analysis, an adaptive filter was applied to reduce the effect of radar-system-generated coherent speckle to produces an image approximating local tone while preserving edge definition. A feature set was than chosen that best classifies the SAR image into the aimed classes. The features are selected based on their discrimination ability and classification accuracy. And at last, the three ANNs used were compared using the image formed by the chosen features in combination.