Aurora Classification in All-Sky Images via CNN–Transformer

Abstract

An aurora is a unique geophysical phenomenon with polar characteristics that can be directly observed with the naked eye. It is the most concentrated manifestation of solar–terrestrial physical processes (especially magnetospheric–ionospheric interactions) in polar regions and is also the best window for studying solar storms. Due to the rich morphological information in aurora images, people are paying more and more attention to studying aurora phenomena from the perspective of images. Recently, some machine learning and deep learning methods have been applied to this field and have achieved preliminary results. However, due to the limitations of these learning models, they still need to meet the requirements for the classification and prediction of auroral images regarding recognition accuracy. In order to solve this problem, this study introduces a convolutional neural network transformer solution based on vision transformers. Comparative experiments show that the proposed method can effectively improve the accuracy of aurora image classification, and its performance has exceeded that of state-of-the-art deep learning methods. The experimental results show that the algorithm presented in this study is an effective instrument for classifying auroral images and can provide practical assistance for related research.