Few-Shot Learning UAV Recognition Methods based on the Multi-Domain Lightweight Network

The widespread use of unmanned aerial vehicles (UAVs) poses safety and management challenges, especially in low-altitude and densely populated areas. Conventional recognition methods have their constraints while radio frequency (RF)-based methods, despite their promise, often face challenges in generalizing effectively when dealing with a small number of samples. In this paper, two few-shot learning UAV recognition methods are proposed. These methods leverage a multi-domain architecture that fully utilizes both time and frequency domain features. By incorporating lightweight network design and attention mechanisms, the proposed methods reduce the number of parameters while maintaining high accuracy. This approach is particularly beneficial for organizations with limited resources or tight time constraints, offering a practical solution for UAV recognition in environments where obtaining a large number of datasets is impractical. Furthermore, simulation experiments show that the proposed method reaches an accuracy of $88.14 \%$ in the few-shot recognition task with only fifteen samples.