An adaptive technique based on separability index to classify PALSAR data

In this paper, we have developed an adaptive technique based on separability index for classification of PALSAR data. First of all this paper investigate the effect of sample size on accuracy assessment. Further, we have utilized minimum distance to mean classifier and studied the sample size varying from 20 to 500. A comprehensive analysis was made for different polarized images, i.e., HH, HV, VV and their combinations i.e., HH-HV, HV-VV, HH-VV, and HH-HV-VV. Further, to increase the classification accuracy, separability index, which tells about the idea of class separation in different polarization was analyzed. A decision tree based on separability index was developed and adaptive statistical approach based on mean and standard deviation was used to classify PALSAR data. Good overall accuracy and class wise accuracy was observed with this adaptive technique.