SAR target recognition using block-sparse representation

We propose a block-sparse representation approach with wavelet approximate coefficients features for synthetic aperture radar (SAR) target recognition. Inspired by sparse representation-based classification (SRC), we take the block structure of the dictionary into account, and introduce block-sparse representation to find one or few blocks in one class to represent the test sample. We use block orthogonal matching pursuit (BOMP) to obtain the linear representation coefficients associated to atoms from one class. Experiments are carried out on Moving and Stationary Target Acquisition and Recognition (MSTAR) public database. We compare our method with SRC. Numerical results demonstrate that the proposed method can improve classification accuracy of SRC and approximate coefficients features perform better than amplitude values features.