Multiview Learning for Automatic Classification of Multiwavelength Auroral Images

Auroral classification plays a crucial role in polar research. However, current auroral classification studies are predominantly based on images taken at a single wavelength, typically 557.7 nm. As a result, the integration of information from multiple wavelengths has received comparatively less attention, resulting in low classification rates for complex auroral patterns. Furthermore, existing studies employing traditional machine learning or deep learning approaches have not achieved an optimal balance between accuracy and speed. To overcome these challenges, this article proposes a lightweight auroral multiwavelength fusion classification network, MLCNet, based on a multiview approach. First, we develop a lightweight feature extraction backbone to improve the classification rate and effectively cope with the increasing amount of auroral observation data. Second, considering the existence of multiscale spatial structures in auroras, we design a novel multiscale reconstructed feature module. Finally, to highlight the discriminative information between auroral classes, we propose a lightweight attention feature enhancement (LAFE) module. The proposed method is validated using auroral observations from the Arctic Yellow River Station (YRS) during 2003–2004. The experimental results demonstrate that the fusion of multiwavelength information significantly improves the auroral classification performance. In particular, our approach achieves the state-of-the-art classification accuracy compared to previous auroral classification studies and outperforms existing multiview methods in terms of both accuracy and computational efficiency.