Sparse representation for hyperspectral image target detection

In recent years, there has been a growing interest in the study of sparse representation. Sparse representation has been adopted in this paper for automatic target detection in hyperspectral image(HSI). We using a novel algorithm, the K-SVD, designing overcomplete dictionaries for sparse representation. The sparse representation is recovered by solving a greedy algorithm called Orthogonal Matching Pursuit (OMP) and it directly determines the class label of the test sample. As the neighboring HIS pixels usually have similar spectral characteristics and usually belongs to the same class, we use the smoothing of HIS pixels to constraint on sparsity and improve the reconstruction accuracy. The simulation results of this study demonstrated that it is effective.