Terrain classification in SAR images using principal components analysis and neural networks

Terrain classification from synthetic aperture radar (SAR) images was performed using various neural network architectures. Several different polarization images were used for the training of the neural networks. A region was selected for each class for training of the classifier. The Karhunen-Loeve transform and parametric modeling were used to extract the salient features of the input in each region and reduce the dimensionality of the feature space. The transformed data were used for training and testing purposes. Simulation results on real SAR images are provided.<<ETX>>