基于交互式多模型快速数据关联的多机动目标跟踪算法研究

发布时间：2018-02-10 22:54

本文关键词： 机动目标跟踪交互式多模型快速数据关联扩大关联门自适应采样周期　出处：《太原理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：机动目标跟踪技术作为近几十年来研究的热点,其在军事和民用领域的作用发挥也是越来越大。在机动目标跟踪技术发展历程中,一方面,各种新的运动模型、滤波算法及数据关联算法相继被提出,大大提高了算法跟踪性能,另一方面,随着各类目标机动性能的提高,对目标跟踪系统的要求也是越来越苛刻。因此,对于机动目标跟踪算法的研究无论在理论研究上还是在实际应用中都有非常重要的意义。本文首先介绍了目标跟踪的组成要素、基本原理以及目标跟踪领域里的一些典型的数学模型,主要包括CV模型、CA模型、Singer模型、“当前”统计模型、CT模型等,并归纳总结了各个模型的应用特点。其次介绍了目标跟踪中的滤波算法,重点研究了卡尔曼滤波(KF)、扩展卡尔曼滤波(EKF)、不敏卡尔曼滤波(UKF),并通过仿真实验比较了三种算法的滤波效果。然后,针对目标跟踪过程中的数据关联问题,研究了最近邻数据关联(NN)、概率数据关联(PDA)和联合概率数据关联(JPDA)。其中,基于贝叶斯理论的JPDA算法被公认为多目标数据关联的经典算法,理论上已解决了数据关联问题,但是由于算法在计算关联概率时,需要对所有可能的目标关联解进行搜索,搜索过程运算量的指数增长效应,会导致算法计算量过大,实时性难以满足,为此本文使用快速数据关联(FDA)算法来解决计算负荷过大问题。作为本文的重点研究内容,文中较为详尽的分析了交互式多模型(IMM)算法的基本原理,针对杂波环境中多机动目标跟踪问题,将IMM与FDA算法进行结合,解决目标量测点迹和目标运动状态的不确定性问题。针对跟踪过程中的目标失跟问题,将扩大关联门应用到此算法中,当目标发生机动,关联门内检测不到有效量测时,通过逐步扩大关联门的方式,重新搜索有效量测,降低机动目标跟踪丢失率。同时,为了尽可能地减小新算法的运算量,运用预先定义采样间隔法自适应调整跟踪采样周期,以此平衡目标跟踪精度与跟踪系统负载。
[Abstract]:As a hot topic in recent decades, maneuvering target tracking technology is playing a more and more important role in military and civilian fields. In the development of maneuvering target tracking technology, on the one hand, there are various new motion models. Filtering algorithms and data association algorithms have been proposed one after another, greatly improving the performance of the algorithm. On the other hand, with the improvement of maneuverability of various targets, the requirements of target tracking system are becoming more and more stringent. The research of maneuvering target tracking algorithm is of great significance in both theoretical research and practical application. This paper first introduces the constituent elements of target tracking. Basic principles and some typical mathematical models in the field of target tracking, including CV model, CA model, Singer model, "current" statistical model and CT model, etc. The application characteristics of each model are summarized. Secondly, the filtering algorithm in target tracking is introduced. This paper focuses on the study of Kalman filter, extended Kalman filter, unsensitive Kalman filter and UKF, and compares the filtering effects of the three algorithms through simulation experiments. Then, aiming at the data association problem in the process of target tracking, In this paper, the nearest neighbor data association, probabilistic data association (PDAA) and joint probabilistic data association (JPDA) are studied. The Bayesian theory based JPDA algorithm is recognized as the classical algorithm for multi-objective data association, and the problem of data association has been solved theoretically. However, when calculating the association probability, the algorithm needs to search all possible target association solutions, and the exponential growth effect of the search process will lead to too much computation and the real-time performance of the algorithm is difficult to satisfy. In this paper, the fast data association algorithm is used to solve the problem of large computational load. As the focus of this paper, the basic principle of the interactive multi-model IMM algorithm is analyzed in detail. In order to solve the problem of multi-maneuvering target tracking in clutter environment, the IMM and FDA algorithms are combined to solve the uncertainty problems of the tracking points and the moving state of the target. The extended correlation gate is applied to this algorithm. When the target is maneuvering and the effective measurement is not detected in the correlation gate, the tracking loss rate of the maneuvering target can be reduced by re-searching the effective measurement by expanding the correlation gate step by step. In order to minimize the computational complexity of the new algorithm, the predefined sampling interval method is used to adjust the tracking cycle adaptively to balance the target tracking accuracy with the tracking system load.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN713

【参考文献】