Fusion of Multi-frequency Polarimetric SAR and LISS-3 Optical Data for Classification of Various Land Covers

Abstract:
A comparative assessment of optical and microwave data is essential to find out the complementarity of data for classification of various land features. In this paper, the classification accuracy of LISS-3 (IRS-P6) optical data over Mumbai is compared with multi-frequency PolSAR data to study the synergy between both the remote sensing techniques. SVM classifier is applied on optical and PolSAR data. Before applying SVM classifier, optical data is registered with PolSAR data. The overall accuracy of optical IRS data (87.31%) is less than that of fully polarized SAR data (93.54%). Optical data cannot differentiate well between grasslands and mangroves and also saltpans and wetlands, whereas PolSAR data gave high classification accuracy. However, forest is well classified with optical data. Since the forest is on hilly area, foreshortening and layover in PolSAR data affect the classification accuracy. It is observed that the combination of both optical and PolSAR gives better results (98.62%).