一种基于多模态OAM波束的目标特征智能识别方法  

An Intelligent Target Feature Recognition Method Based on Multi-mode OAM Beams

作　　者：周宁宁[1] 朱士涛[1] 年毅恒 田春明[1] 张安学[1] ZHOU Ningning;ZHU Shitao;NIAN Yiheng;TIAN Chunming;ZHANG Anxue(School of Information and Communication Engineering,Xi’an Jiaotong University,Key Laboratory of Multifunctional Materials and Structures,Ministry of Education,Xi’an 710049,China)

机构地区：[1]西安交通大学信息与通信工程学院多功能材料与结构教育部重点实验室,西安710049

出　　处：《雷达学报（中英文）》2021年第5期760-772,共13页Journal of Radars

基　　金：国家自然科学基金(62071371,61801368,61801366);超高速电路设计与电磁兼容教育部重点实验室(LHJJ/2020-04);雷达信号处理国防科技重点实验室。

摘　　要：高效率目标探测需要借助探测信号的低相关性空间调制,调制数量大且具有时间独立性。携带轨道角动量(OAM)的涡旋波束具有无穷多种模态且不同模态之间相互正交,借助强色散材料可以实现频率域的多模态OAM波束产生。该文首先对OAM的传播特性进行推导,给出了符合探测需求的多模态OAM波束源特征;在此基础上,研究了不同模态的OAM波束在3种不同应用场景下目标反射回波信号特性,采用卷积神经网络对不同反射场景下的数据特征进行提取,实现了对未知场景的判断及场景内的目标识别,并进行了抗噪性能分析。实验结果表明:理想状态下,网络对目标场景判断的准确率可达97:5%;各反射场景中的两个相邻目标的间隔大于某一阈值时,网络对目标位置的识别准确率均高于80%。但目标识别效果有环境依赖性,当SNR<20 dB时,3种场景内的目标识别准确率均大幅降低。Target detection based on space modulation requires a large number of test modes with space-time independence.The Orbital Angular Momentum(OAM)beams are orthogonal to each other and have infinite modes.Due to the strong dispersive materials,multi-mode OAM beams with the same scattering angle can be generated in the frequency domain.In this manuscript,the propagation characteristics of multi-mode OAM beams are analyzed,which can be utilized to improve detection efficiency.The echoes from the target illuminated by the multi-mode OAM beams are then investigated in three different application scenarios.A convolution neural network is employed to extract the relationship between the echo data and the target image based on prior knowledge.The target and the imaging scenarios can be distinguished with a high probability.Finally,the proposed method’s anti-noise performance is analyzed.The experimental results show that in the ideal state,the accuracy of target scene judgment can reach 97.5%.The accuracy of the target location recognition is higher than 80%when the interval between two adjacent targets in a scene is larger than a threshold.The accuracy of the target location recognition in three scenes is greatly reduced when SNR is less than 20 dB,depending on the scene.

关 键 词：空间调制 轨道角动量(OAM) 多模态涡旋电磁波 卷积神经网络 特征提取 目标识别 

分 类 号：TN95[电子电信—信号与信息处理]