A Data-Driven Approach to Small UAV Detection using Micro-Doppler Signatures and Deep Convolutional Architecture

Abstract - Detecting and classifying Small Unmanned Aerial Vehicles (SUAVs) remains a complex task for radar systems due to their compact size and radar cross-section, which often resemble those of birds, insects, or background clutter. As SUAVs are increasingly used in security-sensitive contexts, improving detection methods is essential for defence and airspace monitoring. Traditional radar approaches often fall short in reliably distinguishing SUAVs from similar targets. However, micro-Doppler signatures derived from Continuous Wave (CW) radar offer valuable motion-based features that can help overcome this challenge. In this study, we propose a method to increase the size of micro-Doppler signatures obtained from Continuous Wave (CW) radar, aiming to enhance the discriminative capabilities of radar-based UAV detection systems by augmenting the collected radar data and employing advanced signal processing techniques. This enhancement captures subtle motion characteristics that are key to differentiating SUAVs. We further leverage recent advancements in Artificial Intelligence (AI), particularly Deep Learning (DL), to enable automatic feature extraction and classification. Our experimental results indicate that the proposed approach improves detection robustness, achieving a classification accuracy and balanced performance across precision, recall, and F1-score. These findings underscore the potential of our method to strengthen radar-based UAV detection systems in clutter and noise environments. Key Words: Artificial Intelligence (AI), Airspace Security, Continuous Wave (CW) Radar, Deep Learning (DL), Micro-Doppler Signatures,