Joint sparse representation of hyperspectral image classification based on quadratic space processing

In view of the problems of low classification accuracy and insufficient utilization of spectral and spatial information in many traditional algorithms applied to hyperspectral image classification, a joint sparse representation hyperspectral image classification algorithm based on quadratic spatial processing was proposed. The morphological features are extracted before the dictionary training, and the initial dictionary is constructed together with the spectral features to achieve the purpose of training higher quality dictionary atoms faster. In order to make full use of spatial information, obtaining edge information through superpixel segmentation firstly, and then adaptively selects neighboring atoms through weight calculation under the dual constraints of superpixel edges and fixed neighborhoods to realize the secondary utilization of spatial information. Simulation experiments on two commonly used datasets prove that the algorithm proposed in this paper can effectively improve the classification accuracy.