Detecting UAVs is crucial for maintaining airspace security, preventing unauthorized aerial intrusions, and safeguarding both privacy and property from potential threats posed by these aerial systems. Effective UAV detection relies significantly on analyzing radio frequency (RF) data. Each UAV emits a unique RF signature, which, when carefully analyzed, enables accurate identification, tracking, and classification of these drones. This capability is essential not only for securing airspace but also for complying with relevant regulatory standards. In this context, our research investigates the use of sophisticated deep learning techniques to improve UAV detection through RF data analysis. Specifically, the study examines three different deep learning models: DNN, CNN, and LSTM networks. The performance of these models is assessed in terms of their ability to detect UAVs, classifying their types, and determining their operational modes. The results demonstrate that among the models tested, the LSTM network exhibits superior performance. It achieves a remarkable accuracy of $\mathbf{99.9} \%$ in detecting UAVs, $\mathbf{92.7 \%}$ in classifying UAV types, and $\mathbf{72.5 \%}$ in identifying UAV modes, surpassing the other architectures in all evaluated metrics.
UAV Detection with Radio Frequency Data and Deep Learning Techniques
2024-10-23
792706 byte
Conference paper
Electronic Resource
English