【作者】 吴耀云；

【导师】 陈卫东；

【作者基本信息】 中国科学技术大学 ， 电磁场与微波技术， 2008， 博士

【摘要】 采用泛光收发波束的宽带多基地雷达系统具有观测视野开阔、系统响应快速、测量精度高、结构简单和成本低等诸多优点,在军事和民用领域具有广阔的应用前景。然而该系统的各接收机通常仅能获取目标在观测视线上的一维信息(例如距离和与速度和信息),因此对目标空间位置和运动状态的测量必须综合多个接收机的信息才能实现,正是这一系统特点给复杂背景中的多目标处理带来了较大的困难。本文针对宽带多基地雷达系统的多目标检测、跟踪以及目标信息融合问题,着重从以下两个方面开展了相关研究:(1)、基于航迹关联思想的多目标回波配对检测和多目标跟踪的研究。针对宽带多基地雷达系统采用的对称三角线性调频连续波(STLFCW)工作体制,本文提出了基于“扫频段局部航迹”关联的多目标回波配对检测算法。该算法充分利用多观测周期中同一目标回波间的相关性,将上升扫频段建立的多目标“局部航迹”与下降扫频段的“局部航迹”进行匹配,显著改善了多目标配对的性能,有效剔除了大量虚警形成的各种虚假航迹,提高了复杂环境中的多目标检测能力。该算法在完成多目标配对检测的同时,还可以获取各目标距离和与速度和的精确估值。针对多基地雷达系统收发分置的结构特征,本文提出了一种分布-集中式的多目标快速跟踪算法。该算法利用各接收机中形成的多目标“局部航迹”,通过构造基于速度和信息的假设检验,对多目标“局部航迹”进行数据关联,分级剔除了可能存在的虚假目标组合。并在此基础上,综合接收机的局部跟踪处理(分布处理)和融合中心的“局部航迹”关联与航迹分配管理(集中处理)等跟踪技术,建立了可行的多基地雷达多目标跟踪实时处理的框架。该算法的核心是利用“局部航迹”关联替代了传统方法中复杂的量测-量测和量测-航迹关联处理,因而可以显著提高多目标跟踪的实时性和稳定性。(2)、基于高分辨距离像(HRRP)的交汇目标检测和多视角距离像融合的研究。针对航迹交汇中的目标可靠检测问题,利用目标航迹信息和距离像历史信息,提出了一种基于模糊C-均值聚类算法(FCM)和距离像重建(RPR)技术的交汇目标检测算法。该算法利用目标航迹的预测信息对FCM算法进行初始化,克服了FCM算法对初始条件敏感的缺陷,改善了杂波环境中的目标检测性能。同时,利用航迹交汇前的目标平均距离像,将交汇中混迭的多目标距离像进行分离和重建,从而极大扩展了多目标的可分区域,提高了航迹交汇区域目标参数提取的可靠性。针对高分辨距离像对目标姿态敏感的问题,提出了基于最大熵准则的多视角距离像信息融合方法,降低了距离像对目标姿态变化的敏感性。该方法以目标跟踪获取的目标状态信息为基础,对同一目标的多视角距离像进行了配准和最优加权融合,给出了最优加权系数的解析表达式及其物理解释。仿真表明,多视角距离像的融合处理提高了目标距离像的稳定性,为基于高分辨距离像的目标特征提取和识别奠定了良好基础。更多还原

【Abstract】 The wideband multi-static radar system with omnidirectional antennae has the advantages of wide observation area,fast response,high measuring accuracy,simple structure and low cost,therefore,there is a promising application for this system in military and civil fields.However,each receiver in this system can only observe one-dimensional target information(e.g.range sum and velocity sum) in the radar LOS(line of sight),so all receivers’ measurements must be integrated to estimate target’s state of three-dimensional position and velocity.Due to this special working mode,the processing becomes fairly difficult when tracking multiple targets in a complex environment.To solve the problems of this radar system in multi-target detection,tracking and information fusion,this dissertation carries out research systematically in following two aspects:(1) The investigation on multi-target pairing/detection and tracking based on local track association.A new multi-target pairing and detection algorithm based on "sweeps’ local track" association is proposed for the system working with symmetrical triangle linear frequency modulated continuous waves(STLFMCW).Based on the correlation of the same target in consequential observation cycles,the "local track" in positive sweep is utilized to match with that of comes from the same target in negative sweep,and the pairing performance has improved significantly.As all false tracks caused by false alarms have been eliminated in paring process,the multi-target detecting performance in dense clutters can be enhanced too.At the same time,the target’s accurate range sum and velocity sum can be estimated when all targets have been paired off and detected.Then,a fast multi-target tracking method with distributed-centralized framework has been devised to accommodate the separated transmitter and receiver structure.In this algorithm,the "local track" formed in each receiver has been fully utilized.A hypothesis testing based on velocity sum information is introduced to remove all false target combinations and "local track" association can be realized.Then,the tracking techniques of local tracking in receivers(distributed treatment),"local track" association and track management in fusion center(centralized treatment) are integrated to build up a feasible multi-target tracking framework with real time performance.Since the "local track" association is adopted instead of complicated measurement-to-measurement and measurement-to-track associations in fusion center,the stability and real time performance can be evidently increased.(2) The investigation on track-crossing targets detection and information fusion of multi-aspect range profiles based on high resolution range profile(HRRP).To solve the detection problem of track-crossing targets,a new track-crossing target detection algorithm based on fuzzy C-means clustering(FCM) and range profile reconstruction(RPR) technique is presented.In this algorithm,the prediction of confirmed target is introduced as the prior information of FCM algorithm to reduce the sensitivity to initialization condition and the detection performance in clutter environment has been improved.At the same time,the average range profile of target before track crossing occurs is used to distinguish and reconstruct range profiles of superposed and inseparable targets.As a result of range profile reconstruction,the discriminable area of track-crossing targets is greatly enlarged and the reliability of state parameter extraction has been increased.To solve the problem that high resolution range profile is very sensitive to target-radar attitude,a fusion algorithm of multi-aspect range profiles is proposed based on maximum entropy principle.On the basis of target state information obtained in target tracking process,the registration and optimal weighted fusion methods are provided for multi-aspect range profiles.Then,the analytical expression of optimal weight value has been deduced and its physical meaning is discussed in detail.Simulations demonstrate that the stability of the range profile to target-radar attitude has been improved by fusing multi-aspect range profiles,which lays a foundation for further processing of feature extraction or target recognition.更多还原