Multi-view SAR Target Recognition Based on Joint Dictionary and Classifier Learning

Sparse representation classification (SRC) has been applied to synthetic aperture radar (SAR) target recognition successfully. In this paper, a new multi-view SRC algorithm based on joint dictionary and classifier learning is developed for SAR target recognition. The proposed method has three prominent characteristics, compared with existing SRC algorithms. Firstly, it is to learn a compact dictionary that can promote representation capability. Secondly, it is capable of improving classification performance by jointly optimizing dictionary and classifier's parameters during training stage. Thirdly, the proposed multi-view method can mine both inner correlation and complementary information for different views of the same target by flexibly selecting representation atoms from the learned dictionary, which can be very useful for scenes where radar's observing angle changes in a wide range. Extensive experiments on the MSTAR database illustrate that the proposed method has better recognition performance than some state-of-the-art methods for SAR target recognition.