基于多标签分类网络的无人机信号检测方法

打开文本图片集

中图分类号：TN974 文献标志码：A

A method for drone signal detection based on multi-label classification network

HE Xihao 1，2 ，QI Liang²，SONG Haiwei，ZHANG Jiawei²，TAO Shifei' （1.School of Electronic and Optical Engineering，Nanjing University of Science and Technology， Nanjing 21OO94，Jiangsu，China;2.Research Center of Advanced RF Chips and Systems，Nanhu Laboratory，Jiaxing 314OOo，Zhejiang，China;3.Nanjing Electronic Equipment Research Institute， Nanjing 21000o，Jiangsu，China）

Abstract：In order to solve the problems such as unbalanced，strong interference and simultaneous existence of multiple frequency-hopping signals in frequency-hopping samples of UAV，an improved frequency-hopping signal detection algorithm based on deep residual network is proposed.The algorithm uses image Mosaic for data enhancement，attention module to improve robustness and accuracy， and adjustment of loss function to alleviate the problem of uneven positive and negative samples in data set.The experimental results show that the absolute matching rate of the algorithm on the multi-UAV signal data reaches 96.46% and the single class classification accuracy is no less than 96.92% ：

Keywords：multi-label classification;anti-UAV；frequency-hopping signals

0 引言

无人机起源于军事领域，随着飞控设备和无线电通信设备的升级，开始往小型化、多元化的方向发展[2]。（剩余7423字）