基于信息几何的雷达目标检测

发布时间：2018-05-26 03:29

本文选题：信息几何 + 目标检测　；参考：《大连海事大学》2017年硕士论文

【摘要】：雷达目标检测的研究一直是十分重要的课题,在民用导航以及军事领域中,人们不断尝试新理论和新方法以提高雷达的检测性能。信息几何是在黎曼流形上利用现代微分几何方法研究统计问题提出的一套新的理论体系,为解决雷达信号处理等问题提供了一种新的思路。本文基于信息几何理论,研究雷达目标信号的检测,把统计检测问题转化为流形上的几何问题求解。本文首先分析了传统雷达目标检测的原理和提高检测性能的方法以及存在的优缺点,说明了将信息几何理论引入的可行性。其次讨论了信息几何理论的一些基本概念,在此基础上通过定义统计流形上两概率分布间的距离,继而建立雷达信号检测的基于KLD和基于测地线的两种距离检测器,并利用信号的多脉冲积累和自身算法优化对这两种距离检测器作了进一步的改进,且讨论了距离检测器在低信噪比下弱目标检测中的应用。仿真结果表明距离检测器的检测性能优于传统的似然比检测且更加可靠。最后在高斯分布杂波下,研究了正定矩阵流形的性质,基于流形上的黎曼几何结构设计了一种矩阵恒虚警检测器的算法流程,并分析了针对非起伏目标和起伏目标不同影响因素变化下的检测门限和检测性能;针对非高斯相关杂波,通过计算流形上由观测数据获得的估计分布与目标有无两种假设分布间距离之差的大小实现目标检测。仿真表明,基于信息几何的矩阵恒虚警检测性能优于常规的单元平均恒虚警检测,且该方法在复杂杂波背景下同样具有较好的检测性能,易于实现,有一定的应用前景。
[Abstract]:The research of radar target detection is always an important subject. In the field of civil navigation and military affairs, people try new theories and methods to improve the performance of radar detection. Information geometry is a new set of theory system proposed to study statistical problems on Riemannian manifold using modern differential geometry method. It provides a new way to solve radar signal processing problems. Based on the theory of information geometry, the detection of radar target signal is studied in this paper, and the problem of statistical detection is transformed into a geometric problem on manifold. In this paper, the principle of traditional radar target detection, the methods to improve the detection performance and the advantages and disadvantages are analyzed, and the feasibility of introducing information geometry theory into radar target detection is illustrated. Secondly, some basic concepts of information geometry theory are discussed, and then the distance between the two probability distributions on the statistical manifold is defined, and then two distance detectors based on KLD and geodesic are established for radar signal detection. Further improvements are made to these two range detectors by using multi-pulse accumulation and self-algorithm optimization, and the application of range detectors in weak target detection with low SNR is discussed. Simulation results show that the detection performance of the distance detector is better than the traditional likelihood ratio detection and is more reliable. Finally, the properties of positive definite matrix manifold are studied under Gao Si distributed clutter. Based on Riemannian geometry structure on manifold, an algorithm flow of matrix CFAR detector is designed. The detection threshold and detection performance of the non-undulating target and the undulating target under different influence factors are analyzed, and the non-Gao Si correlated clutter is analyzed. Target detection is realized by calculating the distance difference between the estimated distribution and the target obtained from the observed data on the manifold. Simulation results show that the performance of matrix CFAR detection based on information geometry is better than that of conventional CFAR detection, and this method also has good detection performance in complex clutter background, and it is easy to be implemented, and has a certain application prospect.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN957.51

【相似文献】