Hyperspectral Image Classification Based on Attentional Residual Networks

Aiming at the problem that the features extracted by the existing convolutional neural Network model are insufficient and the more important information lacks key Attention in the process of hyperspectral image classification, this paper designs an Attention Residual Network (ARN). Firstly, the residual network is used to extract the features of hyperspectral images, and the quality of feature extraction is effectively improved by solving the problem of gradient disappearance and gradient explosion in deep neural network training. Then, the Attention Module (AM) is introduced to optimize the feature extraction process, so that the model can focus on the important regions in the image. This method was tested on the self-collected Herbage (HB) dataset and the public datasets Indian Pines (IP) and Pavia University (PU), and compared with four classical deep learning classification methods. The experimental results show that the proposed method performs best in hyperspectral image classification and greatly improves the classification accuracy.