【作者】 吴伟；

【导师】 王东进；

【作者基本信息】 中国科学技术大学 ， 电磁场与微波技术， 2007， 博士

【摘要】 本文系统地研究了具有合作照射源的一发多收多传感器系统在多目标跟踪应用方面的几个关键问题。首先研究多传感器的“观测值—观测值”关联问题，这里应用多维分配算法对其求解。为了去除杂波的影响，从而降低多维分配问题的规模，本文提出“局部跟踪—融合处理中心合成跟踪”二级处理方案，在各传感器处建立局部航迹，规定只有与确认航迹关联的观测值(AMR)才被传输到融合处理中心。为了进一步提高多维分配算法的计算效率，本文又提出一种簇技术，以减小多维分配问题中需要计算代价系数的候选关联事件的数目。为了实现只向融合处理中心传输AMR数据的目的，本文提出一种能在杂波环境下使用的带有航迹评价功能的局部多目标跟踪算法。为了提高局部航迹更新的处理效率。本文提出一种聚处理技术，它可以将全局“观测值—航迹”关联问题分解成若干个小关联问题。针对多目标局部航迹可能出现的航迹交叉运动情况，本文提出一种Fixed Lag TSDA-AI的多目标航迹平滑算法。这种算法利用观测值在两个扫描周期的关联性以及信号的幅度信息来提高数据关联的精度，同时应用固定延迟平滑算法进一步提高航迹的滤波精度，避免航迹偏差和航迹合并情况的发生。接着依据多传感器扫描周期很小的特点，本文提出一种应用双滤波器的高精度单目标跟踪算法，分析并衡量具体使用EKF、UKF、PF滤波算法时单目标跟踪算法的性能，为系统选择具体的滤波算法提供理论依据。最后在融合处理中心，依据各传感器的AMR，本文提出一种具有航迹质量评估的动态多维分配的多目标跟踪算法，系统依据航迹质量分别对航迹做出确认和删除的判决。为了进一步提高判决的速度和精度，引入信号幅度信息辅助决策，并从航迹起始时刻正确判决的概率和判决平均样本长度两方面定量分析了信号幅度信息对航迹真假判决的改善程度。更多还原

【Abstract】 In this dissertation the key problems of multitarget tracking are systematically investigated in a multi-sensor system which is composed of cooperative transmitter and several receivers.Firstly, the algorithm of multiassignment is analyzed to solve the "measurement-measurement" data association problem. To reduce the scale of the multiassignment problem, a two level processing scheme is proposed, which composes of local-level tracker and fuse central-level composed tracker. In this form, some local tracks are established in local sensors, and only the associated measurement reports(AMR) that are determined to belong to valid tracks are passed to central-level tracker by way of the communications system. To advancing the calculation efficiency of multiassignment algorithm, a cluster method is proposed, which can reduce the number of candidate associated events that need calculate the cost coefficients.Secondly, To achieve the purpose of only transferring AMR to fuse centre, a local multiple target tracking method is proposed, this method has track evaluation function and can be used in noise environment. To advance the local track updating efficiency, another cluster method is proposed which can decompose the whole "measurement -track" data association problem to many small ones. To escape the track biases and track coalescence for closely-spaced tracks, a multitarget track smoothing method is proposed. In this method, the measurements relationship in two scans and the signal amplitude information are used to advance the precision of data association, and then the fixed lag track smoothing method is used to advance the precise of tracks’ state estimation.Thirdly, based on the characteristic of low scan period in multisensor system, a high precise single target tracking method is proposed which composed of two filters. To offer theory basis for system selecting filter method, the performances are analyzed and compared when the concrete filter such as EKF、UKF and PF are used.Finally, in the fuse processing centre, a dynamic assignment algorithm is proposed for multitarget tracking. Based on the tracks’ score, system can make decision for track validation or track deletion in real time. To advance the speed and precise of decision-making, the signal amplitude information is used in the method. Then, quantitative analysis is carried out in two ways: the probability of correct test at the track’s initializing time and the average sample length of track test.更多还原