Classification of Polarimetrie Synthetic Aperture Radar Images Using Revised Wishart Distance

Synthetic Aperture Radar (SAR) is an advanced active radar imaging technology. It uses polarized electromagnetic waves to capture images of the earth surface. In this paper we have proposed polarimetric synthetic aperture radar (PolSAR) image classification technique. Proposed technique uses single hidden layer neural network to achieve classification task. We have proposed linearization model of revised Wishart distance such that it can be used for training the proposed network. It is used along with k-means algorithm to calculate initial weights of network prior to training. Pre-calculating weights helps network to converge quickly with high classification accuracy. Performance evaluation of proposed network is conducted on NASA/JPL Airborne Synthetic Aperture Radar (AIRSAR) data acquired over Flevoland in Netherlands. Proposed network achieves 93.01% overall classification accuracy on Flevoland dataset.