摘要
针对机载火控雷达空空工作模式识别局限性大的问题,从电子情报和雷达告警系统的视角定义了八种工作状态,提出了基于一维卷积神经网络的识别方法。基于不同工作状态的波形特征开发了信号模拟器,构建了全脉冲数据预处理和工作状态自动识别的一维卷积神经网络结构,迭代训练确定最优的网络参数完成状态识别。仿真结果表明:文中提出的识别方法精度高,且对信噪比小、错漏脉冲多的信号适应能力强,具有较强的工程应用价值。
Aiming at the limitation of airborne fire control radar air-to-air pattern recognition,eight working states are defined from the perspective of electronic intelligence and radar warning system,and a recognition method based on one-dimensional convolutional neural network is proposed.Based on the waveform characteristics of different working states,a signal simulator is developed,and a one-dimensional convolutional neural network structure for full pulse data preprocessing and automatic working state recognition is constructed.The optimal network parameters are determined by iterative training to complete state recognition.The simulation results show that the recognition method has high accuracy and strong adaptability to signals with small signal-to-noise ratio and many wrong or missing pulses,so it has strong engineering application value.
作者
马珂
毕大平
周圣栾
何文波
MA Ke;BI Daping;ZHOU Shengluan;HE Wenbo(College of Electronic Countermeasure,National University of Defense Technology,Hefei Anhui 230037,China;The Unit 95793 of PLA,Guiyang Guizhou 550025,China;The Third Military Representative Office of Air Force Equipment Department in Chengdu,CChengdu Sichuan 610036,China)
出处
《现代雷达》
CSCD
北大核心
2023年第3期17-23,共7页
Modern Radar
关键词
机载火控雷达
空空工作状态
一维卷积神经网络
airborne fire control radar
air-to-air working state
one-dimensional convolutional neural network
