5G Spectrum Learning-Based Passive UAV Detection in Urban Scenario

Unmanned aerial vehicles (UAVs) will play an increasingly important role in urban scenario, while the existing UAV detection technologies are inadequate to achieve full-time, large-scale, and passive UAV surveillance. To solve the problem, we propose a fifth-generation (5G) spectrum learning-based UAV detection method, which leverages the distinctive spectral disturbance features under various conditions of UAV presence. To be specific, the synchronization signal blocks (SSBs) are extracted from the captured 5G I/Q signals to reduce the impact of payload data and the flexible 5G frame structure. Then, the UAV detection task is framed as a multi-class classification task, employing a convolutional neural network (CNN) that utilizes the transformed SSBs and their corresponding labels. Experiment results with on-field data validate the effectiveness of the proposed method.