Advanced Radio Spectrum Sensing for UAVs Detection Using Artificial Intelligence Techniques

Abstract

The rapid growth in the use of unmanned aerial vehicles (UAVs) in commercial and military applications, along with the increasing accessibility of these technologies, has created new challenges for critical infrastructure security, airspace protection, and public safety. In this context, the development of effective methods for detecting UAVs has become crucial for security and defense applications. This paper proposes an artificial intelligence-based framework for UAV detection using radio spectrum sensing methods. The training and evaluation pipeline for the AI model uses a custom dataset consisting of 36,000 RF spectrograms in the time-frequency domain. The dataset was divided into two classes: drone, which includes RF signals from 6 commercial UAVs acquired in different operating modes, and no drone, which includes signals acquired in indoor and outdoor environments. The system has been tested in real-world dynamic scenarios at various distances in congested wireless environments characterized by high RF traffic. The experimental results achieved an accuracy of over 94% in real-world operating scenarios. The obtained results show a high level of performance, highlighting the system’s potential for real-world applications in live RF monitoring and UAV sensing.