A deep learning-based unmanned aerial vehicle signal detection method and system

Abstract

This invention discloses a deep learning-based method and system for UAV signal detection, belonging to the field of UAV signal detection. The method first performs basic parameter definition, data reading, and time-frequency transformation on the IQ signal to generate a time-frequency map. The UAV image transmission signal in the time-frequency map is labeled, and a dataset is constructed. A deep learning model containing a multi-scale feature extraction module, an SE module, and deep convolutional layers is trained. Then, the same processing is performed on the target signal to generate a target time-frequency map, which is input into the trained model. The model outputs time-frequency domain location information and confidence data, which are then visualized. This system, corresponding to the above method, improves signal feature extraction capabilities and anti-interference performance through standardized processes and lightweight model design, reduces the number of model parameters, adapts to edge device deployment, and meets the detection needs in complex environments.

Images