一种基于ResNet的雷达弱小目标检测方法  

A Detection Method for Radar Weak Targets Based on ResNet

作　　者：邱明劼 张鹏[1] 汪圣利[1] QIU Mingjie;ZHANG Peng;WANG Shengli(Nanjing Research Institute of Electronics Technology,Nanjing 210039,China;China Academy of Electronics and Information Technology of CETC,Beijing 100041,China)

机构地区：[1]南京电子技术研究所,江苏南京210039 [2]中国电子科技集团公司电子科学研究院,北京100041

出　　处：《无线电工程》2024年第7期1652-1659,共8页Radio Engineering

基　　金：国家自然科学基金(62101261)。

摘　　要：为了解决恒虚警率(Constant False Alarm Rate,CFAR)检测算法对雷达弱小目标检测困难的问题,研究了基于卷积神经网络(Convolutional Neural Network,CNN)的目标检测方法。充分利用神经网络在特征提取上的优良性能,提出了一种基于残差网络(Residual Network,ResNet)块的雷达弱小目标检测方法。突破了传统方法仅利用幅度信息进行目标检测的框架,充分挖掘雷达回波数据中目标的相位特征作为神经网络目标分类检测的依据。经实验验证,所提出的方法在目标回波信噪比仅有-7 dB情况下,仍可实现50%以上的发现概率,并且随着信噪比的降低,所提方法的优异性越发明显。In order to solve the problem that Constant False Alarm Rate(CFAR)detection algorithm is difficult to detect radar weak targets,the target detection method based on Convolutional Neural Network(CNN)is studied.Taking full advantage of the excellent performance of neural networks in feature extraction,a radar weak target detection method based on the Residual Network(ResNet)block is proposed.This method breaks through the framework of traditional methods using only amplitude information for the object detection,and the phase features in radar echo data are fully mined as the basis for neural network object classification detection.According to experiments,the proposed method can still achieve a detection probability of over 50%even when the signal-to-noise ratio of the target echo is only-7 dB.Moreover,as the signal-to-noise ratio decreases,the superiority of the proposed method becomes more apparent.

关 键 词：恒虚警率检测 残差网络 弱小目标检测 

分 类 号：TN959.1[电子电信—信号与信息处理]