Generative adversarial network-based hyperspectral image unsupervised feature extraction method

The invention discloses a generative adversarial network-based hyperspectral image unsupervised feature extraction method. The invention aims to solve the problems of poor feature extraction effect and need for a large amount of label information in the prior art. The implementation process of the method includes the following steps that: data preprocessing is carried out; the parameters of a generative adversarial network are set; the output of a generator network G in the generative adversarial network is calculated; the objective function of the generative adversarial network is constructed; the generative adversarial network is trained; and multi-layer features are extracted, pooling and combining operation is performed on the multi-layer features, so that final features can be obtained. According to the method of the invention, the feature extraction of the generative adversarial network is adopted for the first time; convolution calculation utilizes the spatial information and spectral information of hy