Classification of PALSAR data utilizing class separability index

The prime objective of this paper is to find out the best polarization or polarization combination for classification of Polarimetric Synthetic Aperture Radar (SAR) images. The Polarimetric information contained in Polarimetric SAR images represents great potential for characterization of natural and urban surfaces. In this paper we have used PALSAR image for our objective. Major land covers are Water, Urban and Vegetation. Correlation coefficient is calculated between HH, HV and VV polarized images. HH and VV polarized images have strong correlation coefficient to each other. The concept of separability index has been used for segregation of polarization indices and each class from other classes. Based on image statistics and separability index decision tree classification is developed using best polarization index for each class. Obtained classified image shows good overall and class wise accuracy.