Simulating Meteor Echoes for Advancing All-Sky Meteor Radar Capabilities
We present a machine learning approach to meteor detection using a YOLO-X object detection network trained entirely on synthetic radar data. The model is applied to all-sky meteor radar observations and evaluated against traditional detection methods. Compared to these traditional techniques, the YOLO-X network demonstrates improved sensitivity to low-power meteor echoes and lower rates of false-alarms. These findings suggest that machine learning models trained on synthetic data can improve detection metrics for specular meteor radar systems. Our results support the application of these networks to new and existing radar systems for higher meteor counts, lower error rates, and eventually, more accurate estimation of all measured or inferred meteor characteristics.