Real-time automated detection of multi-category solar radio bursts

Accurate real-time solar radio burst (SRB) detection is crucial for solar physics research and space weather forecasting. Currently, most studies on solar radio burst detection focus on single-category identification and simple discrimination of bursts. There are limited existing studies on multi-category detection. This paper proposes a real-time multi-category solar radio burst detection method to meet the requirements of real-time detection, detection accuracy, and classification accuracy in solar radio bursts. First, solar radio burst spectrums were collected from e-CALLISTO. The spectrums are labeled using LabelImg, and a dataset containing solar radio bursts of Type II, Type III, Type IIIs, Type IV, and Type V was established. Second, a full-dimensional dynamic convolution was introduced in the backbone module of the YOLOv8n model, enhancing the model’s feature extraction capability. Third, a multi-scale feature fusion network based on ConvNeXt was created to prevent feature information loss and optimize the loss function. The experimental results show that the proposed method achieves an average detection accuracy of 82.4% on the established solar radio burst dataset. Compared with the original YOLOv8n model, the accuracy increased by 3.5%. Additionally, the model operates at 140.9 frames per second, with each frame representing a spectrum of 15 minutes duration. Thus, the improved YOLOv8n model enhances the detection accuracy and speed of solar radio bursts, enabling automatic detection and localization of solar radio bursts of Type II, Type III, Type IIIs, Type IV, and Type V.