Probability density functions based study for identification of land cover using SAR data

Fully polarimetric SAR data has the ability of characterizing and differentiating various land covers as it conserves detailed information of the amplitude and the phase of backscattering coefficient, which helps in distinguishing diverse scattering mechanisms. The classification by means of polarimetric data could be enhanced by fusing it with statistical information, but labeling of different classes is still a challenge. So, in this paper, probability density function based approach has been proposed for identification of different classes of land cover. Land cover is classified into four classes using polarimetric indices information and then six probability density functions are applied on each of the classes. Chi-Squared goodness of fit (GoF) test has been used for selecting best-fit density function for each of the classes. The boundaries of the classes were estimated using scale and location parameter of the best-fit density function. The proposed approach was applied on ALOS PALSAR data which resulted in good identification of urban and water region.