【作者】 程之刚；

【导师】 庄钊文；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2007， 博士

【摘要】 本论文研究了警戒雷达组网中的几个关键技术问题。第一章对雷达组网的国内外研究现状、特点和关键技术进行了系统的阐述。第二章研究了警戒雷达组网的布站和数据融合管理的问题。首先提出并定义了布站效率的概念,推导了目标发现概率、虚警率和雷达到目标之间距离的关系式,在此基础上提出布站原则,并且给出了警戒雷达网优化布站的数学模型。为了解决优化布站数学模型中的复杂曲面寻优问题,提出了一种基于多群体搜索的实数遗传算法,该算法具有计算量较小,搜索精度高,收敛速度快,抗早熟能力强,具有一定的搜索多个全局最优点参数的能力,并且非常适合于并行搜索寻优。随后应用布站原则对点目标和给定的RCS曲线目标,使用相同威力范围的雷达进行了纵深部署和前沿部署陆基警戒雷达组网的优化布站仿真研究,给出了有意义的结果。最后应用布站数学模型对给定雷达网覆盖区域、雷达站布放位置和最小目标发现概率的条件下选择雷达的问题进行了仿真计算。在数据融合管理的问题方面,提出了一种具有复杂拓扑结构的数据融合管理网络。分析了实现该网络的关键问题,提出并定义了数据融合子网的性能代价函数,在此基础上给出了选择最优数据融合子网规模的准则,并且进行了仿真实验。第三章分析了警戒雷达组网数据融合的特点。严格地推导了探测数据的坐标转换、探测均方误差的坐标转换和融合权重计算公式。针对雷达网中雷达站布站产生的误差提出了一种基于LMS的雷达网系统误差校正方法。为了适应警戒雷达网目标跟踪的需要,提出了改进的自适应Kalmen滤波跟踪算法,并且为了不同的应用目的还提出了一种SWLMS(短窗LMS)滤波算法。应用仿真实验对所有的算法进行了验证。第四章对警戒雷达组网的目标识别进行了研究。首先分析了应用已知的目标RCS曲线和低分辨雷达进行目标识别所存在的问题。依据信息熵的理论提出了一种将探测数据、RCS样本曲线通过可能性概率进行匹配的目标识别方法。最后进行了仿真实验,仿真实验证明警戒雷达网使用该方法进行目标识别是有效的。第五章总结了本文所做的工作,并且提出了关于雷达组网更深入研究的建议。更多还原

【Abstract】 The intent of this dissertation is to investigate several key technical problems in surveillance radar netting.In the first chapter, there are brief reviews and comments of the research background as well as the present situation of the subject, and then the characteristic feature and key technical problems in radar netting are introduced.The second chapter studies surveillance radar netting and sensor management of data fusion. Firstly, there are deductions of target detection probability, false alarm probability and relation formula of radar-target range. Then the conceptions of station distribution efficiency and distribution principle are proposed and defined. Based on the conclusion we develop an optimization model of surveillance radar netting. Also, a real genetic algorithm based on multi species is proposed for the complicated surface optimization in radar netting model. This algorithm has merits of less computation, higher search veracity, rapider convergence, stronger ability against precocity, and can search multi global optimal points at the same time, and also suits for parallel optimal searching. By applying the station distrbution principle to point targets and targets with given RCS, simulation researches are performed on optimal deploy problems of ground-based surveillance radar net, which contains both depth deploying and front deploying radars with same power range. The simulation shows that this method is feasible and effective. Lastly, based on the radar netting model, simulation calculations are done for the radar choosing under the condition of given radar net cover region, radar station distribution and minimal target detection probability. On the aspect of data fusion management, firstly, a data fusion management network with complicated topological structure is proposed, and the key problems for the network realization are analyzed. Then data fusion subnet performance cost function is proposed and defined. On this basis optimization rule for data fusion subnet size choosing is proposed, and the simulation test is performed with satisfactory results.The third chapter is about characteristic features of data fusion in surveillance radar netting. First, there are rigorous equations for fusion weight calculation, data coordinate transformation and coordinate transformation in detection mean square error. Then considering the system error produced by radar station distribution, radar net system error-correction method based on LMS is proposed. To meet the requirement of target tracking in surveillance radar net, an improved adaptive kalman filtering in tracking calculation is also proposed, as well as an SWLMS filtering algorithm for varied applications. The validity of the presented algorithms is demonstrated with simulated data at last.The fourth chapter focuses on the target recognition in surveillance radar netting. Firstly, existence problems of target recognition based on low-resolution radar and prior target RCS curve are examined. Then on the basis of information intelligence quotient, we develop a new target recognition method, which matches the detective data and RCS sample curve though feasible probability. The validity of the presented method is demonstrated with simulated data at last.The fifth and last chapter concludes the current trends and future outlook in the subject. In particular, some significant and valuable problems of multisensor radar netting are pointed out for further research.更多还原