低截获概率雷达信号侦测与属性识别

The reconnoitered data of low probability of intercept (LPI) radar is generally heavily fragmentary. This key characteristic of the data has caused serious difficulties to intelligence generation. In this project, we intend to innovate array signal processing methods to improve the intercept probability of radar signals, and also innovate data analyzing and radar recognizing methods to improve the recognition probability of radar attributes. The research in this project with do away with traditional radar reconnaissance information processing ideas, and establish basic theory and propose effective methods to improve the intercept and recognition probabilities of various advanced radars, represented by multifunction phased array radars. First, the coherency and temporal continuity of array signals will be made use of, so as to settle the conflict between the instantaneous reconnaissance space and the intercept probability in the energy domain, and enhance the system ability in intercepting weak signals emerging in a large area. Second, we will introduce deep learning techniques to mine the behavior patterns of different radars based on significantly fragmentary reconnaissance data, and formulate automatic recognizers to distinguish between different radars. Through the research of this project, we want to propose feasible signal-intercepting and attribute-recognizing methods for LPI radars in practical environments.

本项目瞄准低截获概率雷达侦察数据的严重残缺这一基本特征，围绕创新阵列信号检测与测向方法提高信号截获概率、创新数据分析与目标识别方法提高雷达属性识别概率两个核心问题，拟突破雷达侦察信息处理的传统思路，针对多功能相控阵等先进体制雷达信号，研究提高其截获与识别概率的基本理论和方法。①结合阵列观测数据的阵元域相干性和时域连续性，调和阵列侦察系统空域与能量域截获概率之间的矛盾，增强对弱信号的瞬时大范围侦测能力；②基于显著残缺的雷达侦察离线大数据，挖掘不同雷达信号参数的时序特征，实现对雷达型号等属性的快速识别。通过项目研究，提出适用于实际环境中低截获概率雷达信号截获和属性识别的有效方法。

本项目瞄准低截获概率雷达侦察数据的严重残缺这一基本特征，围绕创新阵列信号检测与测向方法提高信号截获概率、创新数据分析与目标识别方法提高雷达属性识别概率两个核心问题，突破雷达侦察信息处理的传统思路，提高对多功能相控阵等先进体制雷达信号的截获与识别概率。..通过一年的研究，项目组取得了多项研究成果，在IEEE Trans. Antennas & Propagation、IEEE Trans. Aerospace & Electronics Systems等本领域顶级刊物上发表长文3篇。具体成果包括：①提出了基于子空间追踪的雷达脉冲信号序贯检测方法，实现对阵元域和时域散布信号能量的联合分集，显著提高了对弱信号的发现能力；②提出了基于深度神经网络的阵列信号波达方向估计方法，对未知模型误差具有很强的适应性；③提出了基于递归神经网络的雷达脉冲序列识别和分选方法，能够很好地适应参数联合变化等复杂体制辐射源，以及严重漏脉冲、干扰脉冲等复杂侦察环境，得到了理想的识别和分选性能。..项目还进一步系统研究了对低截获概率雷达目标的时频差定位方法、脉冲分选方法和识别方法，另外撰写了7篇IEEE Trans.论文，目前处于审稿阶段。