Development of a Fast Radio Burst Signal Detection System Utilizing Deep Learning Techniques

Abstract

To detect fast radio burst (FRB) signals from the observational data obtained by FAST radio telescopes, this study develops a recognition system utilizing deep learning object detection algorithms. The system integrates an incoherent achromatization algorithm with the YOLO series target recognition algorithm to identify FRB signals, offering users an intuitive graphical interface. To accommodate varying computational capabilities, the system allows the selection of different algorithm models. Experimental results demonstrate that the system achieves a recall rate of 86% and an accuracy of 83% when tested on the real-world dataset FRB20201124A.