【作者】 赵志超；

【导师】 肖顺平；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2010， 博士

【摘要】 导弹防御是国家安全的重要保证,雷达组网是防空、反导领域的共性问题。导弹防御系统由大量的雷达和红外传感器组成,对弹道目标的预警、跟踪和识别是实现成功拦截的前提条件。本文面向导弹防御技术研究和发展的需求,以雷达网数据融合技术为主线,研究了雷达网的定位、关联、融合、航迹起始等技术,分析了雷达网的定位精度及检测能力,构建了导弹攻防对抗仿真试验系统。主要工作包括以下内容:研究了集中式雷达网定位技术。针对主动雷达网的点迹融合,分析了观测误差随目标位置的动态变化关系,提出了一种动态加权融合算法。针对被动雷达网的测向定位,提出了一种加权最小二乘算法,该算法计算量小、精度高,不需迭代即可逼近克拉美-罗界(Cramer-Rao Bound, CRB),针对单站连续测角情况,推导了其序贯最小二乘实现。分析了雷达布站和雷达性能对雷达网定位精度的影响,推导得到了异类雷达网定位精度CRB的解析表达式,为评估雷达网定位性能提供了一个基准。建立了描述传感器量测误差分布的概率网格模型,并以此为基础提出了联合概率密度矩阵(JPDM)算法。JPDM算法将多传感器定位、关联、融合问题统一转化为图像域中的峰值提取问题,从而避开了关联这个NP难题,该算法本质上是一种极大似然估计,其计算量和定位精度是可控的,并且便于并行处理。提出了基于概率网格Hough变换的多雷达航迹起始算法,解决了常规Hough航迹起始算法中积累单元大小与航迹质量间的固有矛盾。研究了分布式雷达网航迹关联技术。分析了利用航迹曲线宏观特征进行关联的可行性,针对非机动目标,提出了一种基于多项式拟合的异步航迹关联算法;针对机动目标,提出了一种基于B样条拟合的异步航迹关联算法。两种新算法通过参数空间聚类,避免了传统算法的两两关联比较,可以同时处理多条航迹并直接给出融合结果,计算效率显著提高。研究了雷达网的检测与定位性能。建立了单部雷达及雷达网的检测概率计算模型,评估了不同布站下雷达网的综合检测性能。利用微波暗室实测数据分析了雷达网对典型弹道目标的检测能力。开发了基于Visual C++和STK的弹道导弹攻防对抗仿真系统,结合几个典型战情分析了爱国者雷达网、宙斯盾雷达网以及早期预警雷达网对弹道目标的检测与定位能力,得到了一些有意义的结论。论文研究工作来源于实际项目需求,其中的集中式雷达网定位算法和分布式雷达网航迹关联算法已应用于航天工业部门的半实物仿真试验系统,为武器型号的研制与定型提供了有力支撑,基于概率网格的雷达网数据融合算法为构建一体化的战场综合信息感知系统提供了一种可行方案。论文研究成果既可以为我国构建自己的导弹防御系统提供技术储备,也可以为导弹突防策略优化提供参考依据,具有重要的军事价值和现实意义。更多还原

【Abstract】 Missile defense is an important guarantee for national security, and radar networkis a common problem in the field of air defense and missile defense. Missile defensesystemusesalargenumberofradarsandinfraredsensorstoimproveitsdetection,trackingand identification capability, which is the precondition of successful interception. In thebackground of needs for research and development of missile defense techniques, basedon the data fusion technology of radar network, we study on the localization, association,fusion, and track initiation technology of radar network. Also, the localization accuracyand detection capability of radar network are analyzed, and a missile attack and defensesimulation system is built. The major work are as follows:Chapter II studies on localization technology of centralized radar network. Firstly,the dynamic relationship between measurement error and target location is analyzed, anda dynamic weighted fusion algorithm used for plot integration of active radar network isproposed. Secondly, the problem of passive radar network localization is studied, and aweightedleastsquaremethodwithlowcomputationcomplexityandhighprecisionispro-posed, which does not need iteration and can approach to the Cramer-Rao Bound (CRB).For single passive station successively measuring case, the recursive least squares imple-mentation is derived. Thirdly, a general formula for calculating the localization accuracyCRB of heterogeneous radar network is derived, which provides a benchmark for evalu-ating the performance of radar network with different disposition.Chapter III studies on radar network data fusion based on probabilistic grid. At first,the sensor measurement error distribution is analyzed, and a probabilistic grid model isestablished. Then, based on the probabilistic grid model, a general framework for local-ization, association, and fusion is constructed, which called JPDM algorithm. Convertingthe NP-hard association problem to peak extraction problem in image domain, the JPDMalgorithm has a controllable computation complexity and is suit for parallel processing.At last, a Hough transform track initiation algorithm based on probabilistic grid is pro-posed, which resolve the inherent contradition between the size of accumulate cell andthe quality of track.Chapter IV studies on track correlation technology of distributed radar network. Thefeasibility of using the holistic characteristics of tracks for correlation is discussed. For non-maneuveringtargets, apolynomialfittingbasedasynchronoustrackassociationalgo-rithm is proposed. For maneuvering targets, a B-spline fitting based asynchronous trackassociationalgorithmisproposed. Comparedwithtraditionalalgorithms,thecomputationcomplexity is reduced remarkably by using clustering in parameter space.Chapter V studies on detection and localization performance of the radar network.At first, a useful model for calculating the detection probability of radar or radar networkis established, and the detection performance of radar network under different dispositionis evaluated. Then, typical ballistic target RCS data measured in microwave anechoicchamber is used for evaluating detection capability of the radar network. Finally, a bal-listic missile defense simulation system is built based on Visual C++ and STK, with sometypical scenario, the performance of Patriot radar network, Aegis radar network and earlywarning radar network is analyzed respectively.The research of this paper comes from the needs of practical project, most of theresearch results have been applied in practice and achieved good effects. Studying onthe data fusion technology of missile defense radar network has important military valueand practical significance, it can not only provide technical reserves for building our ownmissile defense system, but also give advice for strategies optimization of missile pene-tration.更多还原