Fast Radio Burst Detection Model Based on Improved YOLOV8N

Fast radio bursts (FRBs), as an important research topic in astronomy, have significant scientific value. The Dispersion Measure Time Intensity (DMTI) map is a common type of radio astronomical image in which FRBs typically appear in a bowtie-like shape. To address the need for identifying FRBs in DMTI maps, this paper proposes an improved YOLOv8n-based detection model. In order to reduce the number of parameters of the model and improve the computational efficiency, this paper uses the Ghost Conv module and the C3Ghost module to replace the traditional Conv module and the C2f module, thus optimizing the structure of YOLOv8n. Experimental results demonstrate that the improved YOLOv8n model achieves a recall rate of 95.8% and a precision of 97.6% on the test set, showcasing excellent detection performance.