Hyperspectral Image Classification Using Low-Rank Representation and Spectral-Spatial Information

Classification of hyperspectral images is one of the most important processes on these images. Hyperspectral images are high dimensional, so classification of these images is difficult. Therefore, methods that extract low-dimensional subspace structures from the hyperspectral image are considered. The low-rank representation method can extract the low-dimensional subspace structure in the data. This method considers the global structure of the data. In this paper, to preserve the global and local structure in the data, spares and low-rank representation feature extraction method based on spectral and spatial information is presented. The data structure is better revealed using this model, and the discrimination of the features is increased. In this model, each pixel is expressed by a linear combination of dictionary atoms. In addition, to solve the optimization problem, the alternating direction method of multipliers has been used. The simulation results show that the proposed model has better results than other methods.