| 引用本文: | 唐洪涛，罗忠涛，李史灿，曹 健. 基于原型与主动学习的天波雷达干扰检测方法[J]. 雷达科学与技术, 2023, 21(6): 605-612.[点击复制] |
| TANG Hongtao, LUO Zhongtao, LI Shican, CAO Jian. Interference Detection Method Based on Prototype and Active Learning for Sky⁃Wave Over⁃the⁃Horizon Radar[J]. Radar Science and Technology, 2023, 21(6): 605-612.[点击复制] |
|
| 摘要: |
| 本文研究天波雷达基于距离?多普勒（Range?Doppler, RD）图像的干扰检测问题。在干扰检测过程中，错误检测可能是干扰的漏检与虚警问题，为此考虑采用主动学习方法，将算法模型难以判决的样本由专家标注，并将标注样本加入至训练集中以达到提升检测性能的目的。同时，也需要解决训练集样本的冗余问题，为此使用原型数据学习方法，建立有干扰和无干扰样本数据云，有效地降低训练集样本量。实测数据实验表明，原型方法将初始训练集样本数量降低至23.5%，主动学习方法取得的检测准确率为97.42%，而传统监督学习最近邻方法准确率为87.96%。因此，本文方法能够有效提升天波雷达干扰检测能力，为天波雷达是否需要进行干扰处理与换频检测等工作提供可靠依据。 |
| 关键词: 天波雷达 干扰检测 原型数据 主动学习 RD图像 |
| DOI：DOI:10.3969/j.issn.1672-2337.2023.06.003 |
| 分类号:TN958.93 |
| 基金项目:国家自然科学基金（No.61701067，61702065） |
|
| Interference Detection Method Based on Prototype and Active Learning for Sky⁃Wave Over⁃the⁃Horizon Radar |
|
TANG Hongtao, LUO Zhongtao, LI Shican, CAO Jian
|
|
1. School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065, China;2. Nanjing Research Institute of Electronics Technology, Nanjing 210013, China
|
| Abstract: |
| This paper studies the problem of interference detection based on range?Doppler （RD） image for sky?wave radar. In the process of interference detection, error detection may be the problem of missed detection and false alarm of interference. Therefore, active learning method is considered. Samples that are difficult to be judged by the algorithm model will be labeled by experts, and labeled samples will be added to the training set to improve the detection performance. At the same time, it is also necessary to solve the redundancy problem of training set samples. The prototype data learning methods are used to establish interference sample and non?interference sample data clouds, effectively reducing the sample size of the training set. The real data shows that the prototype method reduces the number of initial training set samples to 23.5%, the detection accuracy rate obtained by the active learning method is 97.42%, while the accuracy rate of the traditional supervised learning nearest neighbor method is 87.96%. Therefore, the proposed method can effectively improve the interference detection capability of sky?wave radar, and provide a reliable basis for whether the sky?wave radar needs to carry out interference processing and frequency switching detection. |
| Key words: sky⁃wave radar interference detection prototype data active learning range⁃Doppler image |
