基于深度残差收缩网络的雷达空中目标识别  

Radar air target recognition based on deep residual shrinkage network

作　　者：尹建国 盛文[1] 蒋伟[1] YIN Jianguo;SHENG Wen;JIANG Wei(Air-Defense Early Warning Equipment Department,Air Force Early Warning Academy,Wuhan 430019,China;Unit 95866 of the PLA,Baoding 071051,China)

机构地区：[1]空军预警学院防空预警装备系,湖北武汉430019 [2]中国人民解放军95866部队,河北保定071051

出　　处：《系统工程与电子技术》2024年第9期3012-3018,共7页Systems Engineering and Electronics

摘　　要：雷达空中目标高分辨距离像(high resolution range profile,HRRP)中往往包含一定的杂波噪声,利用HRRP开展空中目标识别需要重点考虑噪声的影响。针对上述问题,提出一种基于深度残差收缩网络(deep residual shrinkage network,DRSN)的雷达空中目标HRRP识别方法。该网络将深度残差网络、软阈值函数和注意力机制结合起来,采用跨层恒等连接方式,不仅可以避免网络层数过深造成梯度消失或梯度爆炸,从而导致网络学习能力下降的问题,还可以有效过滤掉识别过程中噪声特征的影响,使模型专注于目标区域的深度特征识别,提升强噪声背景下模型的识别能力。实验结果表明,相对于其他常用的深度学习模型,所提方法在各个信噪比条件下,识别效果均有一定的优势,该模型对噪声具有较强的鲁棒性。The high resolution range profile(HRRP)of radar air target often contains a certain amount of clutter noise,and it is necessary to focus on the influence of noise to carry out air target recognition using HRRP.To address the above issues,an air target HRRP recognition method based on deep residual shrinkage network(DRSN)is proposed,which combines deep residual network,soft thresholding function and attention mechanism,and cross-layer identity connection method is adopted.DRSN can not only avoid the problem of gradient vanishing or gradient exploding caused by too deep layers of the network,which leads to the degradation of the learning ability of the network,but also can effectively filter out the influence of noisy features in the recognition process,so that the model can focus on the recognition of deep features in the target region and improve the recognition ability of the model in the strong noise background.The experimental results show that the proposed method has certain advantages in recognition effect under each signal-to-noise ratio condition compared with other commonly used deep learning models,and the model has strong robustness to noise.

关 键 词：空中目标识别 高分辨距离像 深度残差收缩网络 噪声鲁棒性 

分 类 号：TN957[电子电信—信号与信息处理]