基于宽窄带综合的飞机机型识别方法  

Aircraft type recognition method based on wideband and narrowband synthesis

作　　者：夏勇[1,2] 薛娇[1] 汪振亚 田西兰 XIAYong;XUE Jiao;WANG Zhenya;TIAN Xilan(No.38 Research Institute,China Electronics Technology Group Corporation,Hefei 230088,China;University of Science and Technology of China,Hefei 230026,China)

机构地区：[1]中国电子科技集团公司第三十八研究所,合肥230088 [2]中国科学技术大学,合肥230026

出　　处：《空天预警研究学报》2024年第1期1-5,共5页JOURNAL OF AIR & SPACE EARLY WARNING RESEARCH

摘　　要：针对传统单模态雷达飞机目标细粒度识别能力不足的问题,提出一种基于宽窄带综合的飞机机型识别方法.通过卷积网络挖掘宽带一维距离像的局部散射特性,递归神经网络挖掘窄带微动回波不同时刻的上下文特性,达到信息互补的目的.同时,针对两种模态在决策融合时存在强弱不平衡问题,设计了一种自适应决策融合网络,实现了对飞机目标的机型识别.利用七种飞机机型的仿真宽窄带数据进行了实验验证.仿真结果表明,本文方法平均识别准确率为76.7%,较传统方法提升6.4%以上.This paper proposes an aircraft type recognition method based on the fusion of wideband and narrowband data to improve the limited fine-grained recognition capability for aircraft targets of traditional single-modal radar.Firstly,convolutional network is used to mine the local scattering characteristics of wideband one-dimensional range profiles,and then recurrent neural network employed to capture the contextual features of narrowband micro-Doppler signatures at various time points,so as to achieve the purpose of complementary information.At the same time,an adaptive decision fusion network is designed to handle the imbalance between the two modalities during decision fusion,thus achieving aircraft type recognition.Finally,the simulation data of seven types of aircraft targets in wide and narrow bands are used for experimental verification.The simulation results show that the average recognition accuracy of the proposed method is 76.7%,which is over 6.4%higher than that of the traditional methods.

关 键 词：飞机机型识别 宽窄带回波 多模态融合 

分 类 号：TN958[电子电信—信号与信息处理]