| 引用本文: | 齐美彬， 李亚斌， 项厚宏， 杨艳芳， 张学森. 基于自注意力机制的雷达弱目标检测[J]. 雷达科学与技术, 2023, 21(4): 431-439.[点击复制] |
| QI Meibin, LI Yabin, XIANG Houhong, YANG Yanfang, ZHANG Xuesen. Weak Radar Target Detection Based on Self⁃Attention Mechanism[J]. Radar Science and Technology, 2023, 21(4): 431-439.[点击复制] |
|
| 摘要: |
| 对于低信噪比下的弱小目标检测，传统的检测算法采用恒虚警（CFAR）的方式定位目标的位置，但其难以设置合适的阈值，无法很好地应对该类目标的检测。针对上述问题，本文提出VU?Net检测方法。该方法首先对雷达回波数据进行处理，得到目标回波的距离?多普勒（RD）矩阵。然后将RD矩阵输入到所提出的网络框架，通过U?Net实现雷达信号的编码与解码，获取RD矩阵中具有辨别性的深度语义特征，实现逐单元的目标位置预测。同时，该网络中引入自注意力模块实现对雷达信号的关系建模，从而提取更加丰富的目标回波特征，提升网络的抗噪性能。实验结果表明，所提方法的检测性能在低信噪比场景下具有较强的鲁棒性，能够实现对弱目标的有效检测。 |
| 关键词: 弱目标检测 深度学习 自注意力机制 深度语义特征 |
| DOI：DOI:10.3969/j.issn.1672-2337.2023.04.010 |
| 分类号:TN971 |
| 基金项目: |
|
| Weak Radar Target Detection Based on Self⁃Attention Mechanism |
|
QI Meibin, LI Yabin, XIANG Houhong, YANG Yanfang, ZHANG Xuesen
|
|
1. School of Computer and Information Engineering, Hefei University of Technology, Hefei 230009, China;2. The 38th Research Institute of China Electronics Technology Group Corporation, Hefei 230088, China
|
| Abstract: |
| For dim target detection under low signal?to?noise ratio, the traditional detection algorithm uses constant false alarm rate （CFAR） to locate the target, but it is difficult to set an appropriate threshold, which cannot well deal with the detection of this kind of targets. Aiming at the above problems, this paper proposes VU?NET detection method. Firstly, the radar echo data is processed to obtain the range?Doppler （RD） matrix of the target echo. Then the RD matrix is input to the proposed network framework, and the radar signal is encoded and decoded by U?Net to obtain the discriminative deep semantic features in the RD matrix, while the target position prediction is realized unit by unit. At the same time, the self?attention module is introduced into the network to achieve the relationship modeling of radar signals, so as to extract richer target echo features and improve the anti?noise performance of the network. Experimental results show that the detection performance of the proposed method is robust in low SNR scenarios, and it can effectively detect weak targets. |
| Key words: weak target detection deep learning self⁃attention mechanism deep semantic feature |
