一种基于DBSCAN和XGBoost的多模式雷达辐射源型号识别方法

A DBSCAN and XGBoost Based Approach for Multimode Radar Emitter Identification

雷达辐射源型号识别是雷达对抗侦察中的重要环节之一,文中提出一种基于机器学习的多模式雷达型号识别方法。首先,使用DBSCAN生成型号模板库,完成类似传统识别的雷达型号匹配;然后,使用XGBoost训练有监督分类器,进一步识别已知型号雷达。文中对多模式雷达型号识别进行仿真验证,通过在仿真数据上的识别分析表明,本方法能够有效识别不同雷达型号的信号。多模式雷达信号参数具有交错间隔分布的特点,通过与多种基线分类器进行比较,本方法对该类信号具有更好的识别效果。

Radar emitter identification is one of the most important components in radar EW reconnaissance. In this article, a machine-learning based approach is proposed for multimode radar recognition. First, the DBSCAN is applied to generating radar templates and matching radar models, which is similar to traditional identification. Second, the XGBoost is used for training a supervised classifier, which is able to identify known radar models. The evaluation of multimode radar recognition is conducted in simulation experiments. The results demonstrate that the proposed approach is capable of discriminating different radar models on simulated data sets. The parameters of multimode radars have the characteristic of crossing interval distribution. Compared with several baseline classifiers, this approach has better performance for multimode radar emitter identification.