摘要
雷达有机械与相控阵两种扫描体制,一般相控阵雷达远优于机械扫描雷达,因此准确识别扫描体制对威胁评估至关重要。传统体制识别基于专家特征与阈值,需分析大量数据,开发效率与准确性较差。提出两种智能识别方法来解决该问题:(1)以脉冲幅度的一阶差分绝对值直方图为特征,通过支持向量机进行分类识别;(2)建立基于注意力机制的深度卷积神经网络,实现特征的自动学习与扫描体制的识别。实验表明,两种方法均有着良好的准确性,且基于深度神经网络的方法鲁棒性更优。
Radars scan mechanically or electronically,where the latter is usually much better.Being faced with an unknown fighter,recognizing its scanning mode is significant to evaluate the threat level.Traditional methods are based on expert features and thresholds,for recognizing more rapidly and accurate,too much data analysis is required.Two intelligent recognizing methods are proposed.One method is adopting the histogram of the absolute-value of pulse amplitude’s first-order difference as features and support vector machine as the classifier.Attention based deep convolutional neural network,which is capable of learning features automatically,is utilized in the other one.The experiment shows that high accuracy could be achieved by the both methods,while the second approach is more robust.
作者
张文峰
牟皓
赵耀东
顾杰
ZHANG Wenfeng;MU Hao;ZHAO Yaodong;GU Jie(Science and Technology Electronic Information Control Laboratory,Chengdu 610036,China)
出处
《电子信息对抗技术》
北大核心
2022年第2期33-37,共5页
Electronic Information Warfare Technology
关键词
雷达扫描体制识别
机械扫描雷达
相控阵雷达
支持向量机
卷积神经网络
recognition of radar scanning mode
mechanical scanned radar
phased radar array
support vector machine
convolutional neural network
作者简介
张文峰(1990-),男,博士,工程师;牟皓(1986-),男,博士,高级工程师;赵耀东(1986-),男,博士,高级工程师;顾杰(1974-),男,博士,研究员级高级工程师。
