Efficient subspace-based recognition method for SAR targets with similar structure

To identify target for the images obtained by synthetic aperture radar (SAR), those subspace-based methods for automatic target recognition (ATR) usually operate on the range subspace of the training samples. When some targets with similar structures need to be distinguished, the corresponding templates are poorly separable because their range subspaces of each other have a big intersection. A method is proposed for SAR target ATR in this paper, which chooses the orthogonal subspace of the sample range subspace as the projection subspace. Consequently, the difference, between the projections of the different type targets on the projection subspace, is enlarged, which significantly improves the identification effect. Experimental results show that the proposed method is superior to the other similar methods.