Few-Shot Learning UAV Recognition Methods Based on the Tri-Residual Semantic Network

Abstract

Unmanned aerial vehicle (UAV) recognition is of increasing importance since UAV is widely applied and imposes threats to the public safety. Although many UAV recognition methods based on deep learning have been proposed by using the radio frequency fingerprints, they depend on a large amount of training samples or have poor performance when the training samples are few. In this letter, in order to tackle those issues, two few-shot learning UAV recognition methods are proposed based on our designed tri-residual semantic network. Moreover, our proposed tri-residual semantic network not only can extract different levels of the feature information, but also can significantly suppress the effect of interference and noise. Simulation results demonstrate that our proposed methods are superior to the benchmark few-shot learning schemes in terms of the recognition accuracy.