基于残差神经网络的无人机射频信号识别

UAV RF Signal Identification Based on Residual Neural Network

针对民用小型无人机对低空领域安全造成威胁及无人机视觉上识别难的问题,提出了一种无人机射频信号识别的方法,使用深度学习技术学习无人机与控制器之间的射频信号特征来识别无人机。首先将射频信号数据集中的无人机信号进行预处理操作,接着使用残差神经网络进行模型训练,最后使用训练好的网络模型对无人机信号进行识别验证。实验结果表明,该方法识别无人机是否存在的准确率达到99.8%,识别无人机型号的准确率达到91.1%,识别无人机运行模式的准确率达到70.3%,且该方法具备较强的鲁棒性和环境抗干扰能力,性能明显优于基准方法。

In order to solve the problem of small unmanned aerial vehicles(UAVs)threating low-altitude area safety and difficulty in UAV visual identification,a method of UAV signal recognition is proposed.The characteristics of radio frequency(RF)signal between the UAV and its controller are learned with deep learning technology to identify the UAVs.The UAV signal in the RF signal dataset is preprocessed and then fed into residual neural network for model training.Finally,the trained model is used to identify and verify the UAV signal.The experimental results show that the accuracy of the proposed method is 99.8%for identifying the existence of UAVs,91.1%for identifying UAV models,and 70.3%for identifying UAV operating modes.The method has strong robustness and environmental anti-interference ability,and its performance is significantly better than that of the benchmark method.