Unsupervised classification for polarimetric Synthetic Aperture Radar image using the fuzzy possibilistic C-means clustering

The polarimetric Synthetic Aperture Radar (POLSAR) image data has the problems of the noisy pixels and vague category boundaries because of its complex scattering mechanism and statistical property, which strongly influence the classification quality, while the fuzzy possibilistic C-means (FPCM) is robust in detecting the noisy pixels and modeling the uncertainty. Hence, we tried FPCM algorithm combined with the four scattering features (entropy (H), anisotropy (A), scattering angle (α) and total power (SPAN)) to classify the POLSAR data. The feasibility of this approach was tested on the JPL/AIRSAR POLSAR data, and the experiment result shows that the clustering algorithm can perform the classification more effectively in contrast to its counterparts FCM and PCM clustering algorithms.