Unsupervised Polarimetric SAR Image Classification Using \mathcal ^ Mixture Model

This letter proposes a polarimetric synthetic aperture radar image classification method based on the expectation-maximization algorithm. It is an unsupervised algorithm that determines the number of classes in the scene following a top-down strategy using a covariance-based hypothesis test. A G p 0 mixture model is used to describe multilook complex polarimetric data, and the proposed algorithm is tested in simulated and real data sets obtaining good results. The classification performance is evaluated by means of the overall accuracy and the kappa indices obtained from the Monte Carlo analysis. Finally, the results are compared with those obtained by other classic and recently developed classification algorithms.