机动目标跟踪与多目标互联算法研究

发布时间：2018-02-09 16:16

本文关键词： 机动目标跟踪强跟踪非线性滤波数据关联　出处：《电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：无论是在军事领域或民用领域,目标跟踪理论及数据关联算法的研究都有十分深远的意义。随着科技的不断发展,各种新的技术手段被应用到目标跟踪技术中来,但是应用环境也越来越复杂,如何快速提高目标跟踪算法及数据关联算法的性能成为亟待解决的问题。滤波算法与数据关联算法是机动目标跟踪中的核心和难点,本文着重对这两方面进行了研究。首先,本文对机动目标跟踪理论的基本原理及组成部分进行了介绍,又分别对机动目标的运动模型和一些非线性滤波器展开了讨论,重点介绍了几种经典的模型及非线性滤波算法,并提出了两种基于强跟踪的非线性滤波改进算法。这两种算法分别针对容积卡尔曼滤波器与平方根不敏卡尔曼滤波器,将强跟踪滤波因子引入到算法中,在保证算法原有特点的基础上改善了鲁棒性差的问题,使这两种算法在面临强机动的时候具有实时调整能力。最后,分别通过两种不同的仿真环境对这两种改进算法进行了仿真分析,仿真结果进一步验证了算法的有效性。其次,本文对于多目标数据关联的特点及任务进行了简要描述,针对多目标数据互联算法中的一些经典数据关联算法(例如最近邻域法、概率数据互联算法、广义数据关联算法、基于交互多模型的概率数据互联算法等)进行了详细研究,文中详细介绍了各种经典数据关联算法基本原理及推导步骤,并对各种算法性能进行了分析比较。最后采用对单目标数据关联算法及多目标数据关联算法分开仿真的方法,将单目标数据关联算法(概率数据互联算法、基于交互多模型的概率数据互联算法)应用于一个结合匀速与匀加速的目标上。仿真结果表明,当目标的运动不能用单一模型来描述的时候,基于交互多模型的概率数据互联算法性能要优于概率数据互联算法。对于多目标数据关联算法采用一个二维平面上交叉运动的两目标仿真环境,对广义概率数据关联算法及联合概率数据关联算法的性能进行了比较,仿真结果表明,GPDA算法的跟踪误差低于JPDA。
[Abstract]:The research of target tracking theory and data association algorithm is of great significance both in military and civilian fields. With the development of science and technology, various new techniques have been applied to target tracking technology. However, the application environment is becoming more and more complex, so how to improve the performance of target tracking algorithm and data association algorithm becomes an urgent problem to be solved. Filtering algorithm and data association algorithm are the core and difficulty of maneuvering target tracking. Firstly, the basic principle and components of maneuvering target tracking theory are introduced, and the motion model and some nonlinear filters of maneuvering target are discussed respectively. In this paper, several classical models and nonlinear filtering algorithms are introduced, and two improved nonlinear filtering algorithms based on strong tracking are proposed. The two algorithms are for volumetric Kalman filter and square root insensitive Kalman filter, respectively. The strong tracking filter factor is introduced into the algorithm, and the problem of poor robustness is improved on the basis of guaranteeing the original characteristics of the algorithm, which makes the two algorithms have the ability to adjust in real time when facing strong maneuverability. The two improved algorithms are simulated and analyzed by two different simulation environments. The simulation results further verify the effectiveness of the algorithm. Secondly, the characteristics and tasks of multi-objective data association are briefly described in this paper. In this paper, some classical data association algorithms (such as nearest neighborhood method, probabilistic data association algorithm, generalized data association algorithm, probabilistic data association algorithm based on interactive multi-model) are studied in detail. In this paper, the basic principle and derivation steps of various classical data association algorithms are introduced in detail, and the performance of these algorithms is analyzed and compared. Finally, the method of separate simulation of single object data association algorithm and multi-objective data association algorithm is adopted. The single objective data association algorithm (probabilistic data association algorithm, probabilistic data association algorithm based on interactive multi-model) is applied to a target with uniform velocity and uniform acceleration. The simulation results show that, When the movement of the target cannot be described in a single model, The performance of probabilistic data association algorithm based on interactive multi-model is better than that of probabilistic data association algorithm. The performance of generalized probabilistic data association algorithm and joint probabilistic data association algorithm are compared. The simulation results show that the tracking error of GPDA is lower than that of JPDA.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN953

【参考文献】