Land Cover Classification Using Dual-Polarimetric SAR Data

Land use land cover (LULC) classification is critical for management and monitoring of the land surface. Various classification methods, varying from pixel-based, object-based, supervised, unsupervised, machine learning algorithms, etc., are being used with the abundant availability of Sentinel-1 dual-polarized SAR data. However, most of these methods require ample feature space or are based on sophisticated classification methods, increasing computational complexity. Having this in mind, we propose two new parameters R_1 and R_2 derived from dual-polarized SAR data for LULC classification. It is interesting to note that various land cover clusters are distinctively separable in the R_1 and R_2 feature space. These two parameters are then used for land cover classification using the simple K-means unsupervised classifier. The classification results are qualitatively assessed with the Google Earth images.