New AI system detects drones using radio signals instead of cameras
Researchers have developed a deep learning system that spots unmanned aircraft by analyzing wireless signals rather than requiring dedicated radar or camera networks. The approach achieves over 99% accuracy in real-world tests and could dramatically cut costs for airports, urban authorities, and enterprises managing airspace security.
Originaltitel: A SDR System for Passive UAV Detection with Deep Learning Method
<p>Unmanned aerial vehicles (UAVs) have emerged as an important tool for communication research in recent years. However, they introduce new challenges for modern urban management. Conventional UAV detection methods, which rely on either vision-based systems or dedicated sensor networks, incur significant deployment complexity and high maintenance costs. To address these challenges, this paper proposes a communication signal-based UAV detection system. Given the poor signal quality passively scattered by the UAV, we employ signal processing techniques to enhance the feature extraction, while implementing customized modifications to the model architecture to accommodate the characteristics of complex-valued inputs. To validate our approach, we conducted comprehensive tests using a software-defined radio transceiver system constructed by USRP-2974 devices. Experimental results demonstrate that the proposed method outperforms other methods and achieves a detection accuracy surpassing 99% in real-world environments.</p>