Aiming at the problem of classification difficulty caused by hyperspectral image information redundancy

limited label samples and difficulty in extracting spectral and spatial information

a hyperspectral image classification algorithm based on deep learning is proposed. The problem of low classification precision in the case of limited samples is solved by generative adversarial network for data augmentation

and the over-fitting of classification network caused by too few training samples is avoided. An efficient three-dimensional（3D）multichannel convolutional neural network is designed to extract spatial-spectral joint information of hyperspectral images and mine deep features for classification. The experimental results show that the proposed algorithm can adaptively extract the spatial-spectral joint features of hyperspectral images with higher classification precision

and the algorithm has strong robustness and network generalization ability

and the effectiveness and feasibility of the algorithm are fully verified.