SAR Image Target Recognition based on Sparse Representation and Cascade Dictionary

It is necessary to recognize the target in the situation of military battlefield monitoring and civilian real-time monitoring. In this paper, a method based on sparse representation and cascade dictionary is presented for SAR image target recognition. The proposed method extracts the generalized two-dimensional principal component analyze features of the training samples to construct sub dictionaries, each sub dictionary can be considered a binary classifier, all the sub dictionaries in sequence arrangement to form a strong cascade classifier. Sparse representation coefficients of the testing samples over the sub dictionaries are computed with the orthogonal matching tracking algorithm. Recognition is realized according to reconstruction error. The experimental results suggest that the new approach can generate ensembles that outperform traditional classifiers and the other sparse representation classifiers which dictionary directly constructed by all training samples or all training samples' feature information in terms of recognition accuracy and recognition time.