Automatic SAR Target Recognition Using Random Subspace Ensemble Classifiers

Automatic Target Recognition (ATR) has become a widely researched problem due to its applicability in domains such as surveillance systems. It can be implemented on various types of images captured by different sensors. In this paper a novel framework for Automatic Target Recognition (ATR) in Synthetic Aperture Radar (SAR) imagery using Ensemble Classifier is presented. A combination of Principal Component Analysis (PCA) and Non-negative Matrix Factorization (NMF) are used as features to a Random Subspace Ensemble with k-NN as base classifiers. The Random Subspace ensemble offers an elegant approach to feature selection when dealing with high dimensional feature set such as in the present case. Our approach has been benchmarked using the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset and results indicate our method outperforms other the state-ofthe-art SAR ATR techniques reported in the literature.