Small sample hyperspectral image classification method based on deep learning joint model

The invention discloses a small sample hyperspectral image classification method based on a deep learning joint model. The method comprises the following steps: firstly, constructing a three-dimensional channel attention convolutional sub-network, designing a three-dimensional channel attention mechanism and fusing the three-dimensional channel attention mechanism into a convolutional neural network, and extracting deep spectral features with discriminability in a hyperspectral image; secondly, constructing a residual multilayer graph convolution sub-network, and extracting a long-distance spatial relationship in the hyperspectral image by constructing multilayer residual connection, thereby learning to obtain richer depth spatial features of the image; and finally, the two sub-networks are combined, and accurate classification of the hyperspectral image is realized through full connection layer combination and feature conversion fusion. The deep learning joint model provided by the invention can fully extrac