Sar polarimetry and optical image analysis for improved land cover mapping

Abstract

Land Use / Land Cover (LULC) is one of the essential information for environmental and socio-economic analysis. It is dynamic in nature and requires continuous assessment. Satellite remote sensing is a powerful tool which provides accurate and detailed observation of land cover information at various spatial and temporal scales. The spectral-based classifications of optical image may lead to inaccurate representation of Earth surface due to cloud cover, mixed pixels, and spectral confusions among the classes. Therefore, additional information derived through Gray-Level Co-occurrence Matrix (GLCM) textural bands, Synthetic Aperture Radar (SAR) backscattering image, and decomposition techniques are required to bring down the limitations and improve the classification accuracy. In this context, the objectives for the research were defined based on the limitation of optical images and to improve the classification accuracy of optical and SAR image. The objectives of this research is to: (i) study the sensitivity of different polarization combinations of SAR data for land cover classification; (ii) reduce the limitations and to improve the classification accuracy for multispectral optical images using SAR dual polarimetric data and GLCM technique; (iii) compare hybrid decomposition techniques of SAR polarimetric data with GLCM technique to improve the classification accuracy and reduce the limitation of a multispectral optical image; (iv) improve the classification accuracy using quad polarization SAR data with decomposition techniques; (v) extract land cover information with a specific emphasis on an agricultural area in a desert region using quad polarization SAR data with decomposition techniques. newline