【作者】 林岳松；

【导师】 钱积新；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2003， 博士

【摘要】 多传感器多目标跟踪与数据关联是多传感器信息融合理论研究的核心内容，主要包含了数据关联和目标跟踪两部分内容。由于在军事技术和民用技术有着广泛的应用，吸引了众多研究者的兴趣，目前该理论的难点是需要在可容许的时间内，密集虚警环境下完成对多个目标的连续跟踪任务。 无源探测传感器具有不易被敌发现的特点，但单个传感器不能实现对运动目标的完全观测，需要利用多个传感器信息来实现对目标的完全观测。如果传感器探测到的是声音等具有较大时间延迟的信息，则需要进一步对数据完成校准。这些因素导致多运动目标无源跟踪与数据关联问题变得异常复杂困难，许多适用于多运动目标有源跟踪与数据关联的算法在这个问题上变得束手无策。 本文针对无源声音探测网络预警系统，研究了多运动目标无源跟踪与数据关联问题中的相关算法，主要贡献与创新在于： 1．从整体上描述了无源声音探测网络的研究背景、意义、基本框架和研究方法，概述了目标跟踪与数据关联的基本理论与方法，重点分析了几种典型的数据关联方法，包括最近邻方法、概率数据关联滤波器(PDAF)、联合概率数据关联滤波器(JPDAF)、多假设跟踪(MHT)以及多维S-D分配算法。 2．针对实际无源探测网络中存在同一探测区域内只布置了一个无源声音探测传感器站的情况，给出了一种单静止站单目标无源纯方位定位与跟踪的算法，提出了一种简单的单静止站多目标无源纯方位定位与跟踪的算法，从而将单静止站多目标问题转化为单静止站单目标问题。 3．利用最小二乘理论，研究了多静止站单目标无源纯方位定位方法，给出了多静止站单目标的视线交叉定位算法，同时针对声音传播的延时特性，提出了处理具有延时特性的最小二乘迭代视线交叉定位算法。 4．研究了多静止站多目标无源纯方位跟踪与数据关联问题，提出了筛选矩阵的概念，使得无源声音探测网内多目标纯方位跟踪与数据关联问题转化为同一探测区域内的多静止站多目标无源纯方位跟踪与数据关联问题；还提出了一种适合于实际工程应用的时空联合数据概率关摘要联算法，该算法解决了无源多传感器多目标跟踪的难题。针对实际工程中存在含有范数有界的噪声统计特性未知的多传感器融合估计问题，提出了分别基于离散代数Riccati方程(DAR丑)和线性矩阵不等式(LMI)的多传感器数据融合系统H二滤波器设计方法。针对一般实际工程系统所具有的特点:1)过程噪声及测量噪声统计特性未知但能量有限;2)含有范数有界的不确定参数。提出了一种新的基于线性矩阵不等式(LMI)的鲁棒从滤波器设计方法。最后简单回顾了全文的工作，并对今后的研究进行了展望。更多还原

【Abstract】 Multisensor-multitarget tracking and data associationarethe kernel content of multisensor-multitarget data fusion, which includes two parts: target tracking and data association. This technology is widely applied in the field of military and civil, many people are interested in it. It is a difficulty that continuous tracks multiple targets with dense ghost in admissible time.Passive detect sensor has the advantage which can’t easy to be detected by enemy, but single sensor can not get all parameters of moving targets. Multisensor’s information is needed to obtain all parameters of move targets. If the information has the time-delay characteristic such as sound, it is needed to adjust all the data from sensors. These factors make the problem of multiple moving targets passive tracking and data association become especially difficulty, many algorithms become helpless in this problem which are available in active multisensor-multitarget tracking and data association.This paper studies on algorithms of multiple moving targets passive tracking and data association in the field of passive acoustic detective network forwarning system, the major contributions and innovations are as following:1. In this paper, We describe the study background, meaning and methods of passive acoustic detective network, summarize the basic theories and methods of target tracking and data association, analyze some tipical data association algorithms include the Nearest Neighbor algorithm（NN）, Probabilistic Data Association Filtering（PDAF）, Joint Probabilistic Data Association Filtering（JPDAF）, Multiple Hypothesis Tracking（MHT）, and Multidimensional S-D Assignment algorithm.2. In detective network, sometimes a surveillance region have only single sensor. In this paper, there are presented that a single stationary station single target passive bearings-only tracking and data association algorithm and a simplified single stationary station multiple target passive bearing-only tracking and data association algorithm, then a single stationary station multiple target tracking problem can be regarded as asingle stationary station single target tracking problem.3. Based on least square method, this paper presents the line-of-sight location method of multiple stationary station single target. Considering the time delay of sound transfer, a least square line-of-sight location method is presented in this paper.4. Filtration matrix is developed to make the problem of multiple target tracking in net can be regarded as a problem of multisensor-multitarget tracking in a same region. A time and space joint probabilistic data association algorithm is developed to solve the difficult problem of passive multisensor-multitarget tracking.5. In many data fusion systems, process noise and/or measurement noise may have unknown statistics characteristics but limited power. In this paper weobtain a characterization of all H_∞ fusion filters based on DAREs and LMIs, respectively.6. This paper is concerned with the problem of H_∞ -norm and thevariance-cost-guaranteed filter for uncertainty systems, in which 1) the process noise and measurement noise have unknown characteristics but limited power; 2) the state and measurement matrices have time-varyingnorm-bounded parameter uncertainty. A new robust H_∞ filter based onLMI approach has been developed to solve the above problem.7. Finally, a brief review of this paper is given, and the future research directions are proposed.更多还原