Radio Frequency Signal Identification for Unmanned Aerial Vehicles Using Deep Learning Techniques

The technology of Unmanned Aerial Vehicles (UAVs) has been remarkably advanced in recent years, which has led to its widespread application across various fields. However, the frequent occurrence of "black flight" incidents poses significant threats to public safety. Consequently, the rapid and accurate detection and identification of UAVs have become a focal point of current research. Utilizing the radiofrequency (RF) signals of UAVs for identification is a common approach. This paper draws upon the cutting-edge achievements of deep learning in the field of image processing. Initially, we extract the time-frequency domain features from the UAV RF signals, transforming the one-dimensional RF signals into two-dimensional time-frequency diagrams. Subsequently, we employ an improved Yolov8 model for training and apply the trained model to the identification of UAV signals. The experimental results demonstrate that this method effectively detects and classifies UAVs by categorizing and identifying 480 time-frequency diagrams. Specifically, it achieves a precision of 1, a recall of 1, and a mean Average Precision (mAP50) of 0.995 at a 50% Intersection over Union (IoU) threshold.