| 摘要: |
| 雷达目标检测技术能够判断回波信号中目标存在与否，并提取目标位置信息。随着雷达图像质量的提升和人工智能技术的发展，利用雷达图像数据通过深度学习方法实现雷达目标检测功能成为一种新的思路。该文首先从雷达目标检测原理入手，对传统和现代两类检测方法进行了梳理，分析了各类检测方法的特点及适用性。然后针对现代雷达回波信号复杂性增大导致传统检测方法统计建模难的问题和机器学习方法特征提取难度大的问题，对深度学习目标检测方法进行了归纳，主要从深度学习算法、雷达回波图像数据类型和应用场景三个方面进行总结。最后分析了深度学习在雷达目标检测应用中面临的挑战，展望了未来的发展趋势。 |
| 关键词: 雷达 深度学习 目标检测 卷积神经网络 |
| DOI：DOI:10.3969/j.issn.1672-2337.2022.06.001 |
| 分类号:TN957 |
| 基金项目:全军军事类研究生重点资助课题（No.JY2020B150） |
|
| Review on Applications of Deep Learning in Radar Target Detection |
|
SHI Duanyang， LIN Qiang， HU Bing， ZHANG Xinyu
|
|
1. Air Force Early Warning Academy, Wuhan 430019, China;2. Unit 95174 of PLA, Wuhan 430040, China;3. Unit 63650 of PLA, Heshuo 841700, China
|
| Abstract: |
| Radar target detection technology can judge whether the target exists in the echo signal and extract the target position information. With the improvement of radar image quality and the development of artificial intelligence technology, using radar image data to realize radar target detection through deep learning method has become a new idea. In this paper, the principle of radar target detection is firstly discussed, the traditional and modern detection methods are sorted out, and the characteristics and applicability of various detection methods are analyzed. Then, aiming at the problem that the increased complexity of modern radar echo signal leads to the difficulty of statistical modeling of traditional detection methods and the problem that feature extraction of machine learning method is difficult, the deep learning target detection methods are summarized, mainly from three aspects as deep learning algorithm, radar echo image data type and application scenario. Finally, the challenges faced by deep learning in radar target detection applications are analyzed, and the future development trends are prospected. |
| Key words: radar deep learning target detection convolutional neural network |
