SAR Image Target Detection Method Based on Sparse Representation

Abstract

An image target detection algorithm based on sparse representation theory is proposed in this paper for the problem of SAR target detection platform, such as noise, geometric distortion and cover effect. This algorithm is constructing an over-complete dictionary by using SVD Dictionary Learning Method based on the idea of vector clustering. With the completeness of the dictionary, important information in the image can be captured by extracting a small amount of image features, which makes the complex background noise of SAR images generated in the high-speed motion state more robust. On the other hand, the algorithm has the ability of reconstruction, and can reconstruct the test sample by using the sparse coefficient selected by the characteristic function, and select the class with the least reconstruction error as the category of the test sample. The algorithm occlusion has good robustness in target occlusion, rotation and complex background. The detection accuracy and computing speed have improved compared to similar classic algorithms.