基于微多普勒效应特征的空中目标识别  

Air Target Recognition Based on Micro-Doppler Effect Features

作　　者：孟凡君 杨学岭 吴鑫 管志强 MENG Fanjun;YANG Xueling;WU Xin;GUAN Zhiqiang(No.8 Research Academy of CSSC,Nanjing 211153,China;Nanjing University of Aeronautics and Astronautics,Nanjing 211153,China)

机构地区：[1]中国船舶集团有限公司第八研究院,江苏南京211153 [2]南京航空航天大学,江苏南京211153

出　　处：《舰船电子对抗》2023年第4期60-65,共6页Shipboard Electronic Countermeasure

摘　　要：特征提取是基于微动特征雷达目标识别的关键一环。传统方法提取的特征为线性、浅层的,导致表征微多普勒效应的能力有限。针对这些问题,采用非线性网络进行特征学习,建立了基于微多普勒效应的飞机目标识别深度网络。通过构建处理微多普勒效应的卷积神经网络(CNN)模型,从微多普勒频域数据中自动提取非线性深层次属性特征,实现空中目标分类识别。在实际测量的微多普勒频域数据上的大量实验结果表明,所提方法具有良好的目标识别性能和泛化性能。Feature extraction is the key technique for radar target recognition based on micro-Doppler effect.The features extracted by traditional methods are linear and shallow,which results in the limited capability to characterize micro-Doppler effect.Aiming at these issues,this paper a-dopts nonlinear network for feature learning,builds up deep network for aircraft target recognition based on micro-Doppler effect.By constructing a convolutional neural network(CNN)model for dealing with micro-Doppler effect,the nonlinear deep property features of targets are fully extrac-ted from micro-Doppler data of frequency domain,so air target classification and recognition are re-alized.The extensive experimental results on the real measured micro-Doppler frequency domain data show that the proposed method achieves good target recognition performance and generaliza-tion performance.

关 键 词：微多普勒效应 特征提取 雷达自动目标识别 卷积神经网络 

分 类 号：TN957.51[电子电信—信号与信息处理]