Classification of Spectrum Images of Solar Radio Bursts Based on Deep Learning

Violent solar eruptions, such as solar flares and sundial jets, will have a serious impact on the Earth. Not only does it disrupt communication facilities and navigation equipment, but also produce a large number of radio phenomena. Further research on the solar radio phenomena will help us to understand of the activities of solar energy. With the rapid development of deep learning, the feature extraction and classification identification of solar radio spectrum images based on deep learning technology have been widely used. In this paper, gamma transformation, channel normalization, gray stretching, background suppression, image denoising and other image processing techniques, are used to preprocess the spectrum image of solar radio burst. Then we use CNN model, CNN-Capsule model and CNN-LSTM model to research the classification of spectrum images of solar radio burst types. The experimental results show that our approach can effectively improve the classification effect of solar radio burst types.