基于粒子滤波算法的无线传感器网络目标跟踪研究

发布时间：2018-03-28 05:20

本文选题：无线传感器网络　切入点：时钟同步　出处：《华北电力大学(北京)》2017年硕士论文

【摘要】：目标跟踪技术一直以来都是探索的热点,其应用范围涉及军事、工业、商业、医疗等各个领域。目标跟踪技术出现至今已有五十余年,带动了大量的理论与相关技术的进步。随着传感器技术近些年的飞速发展,无线传感器网络出现得到了各方的高度关注,由于其覆盖范围广泛,成本低廉以及检测方式的多样性的特点,无线传感器网络将成为二十一世纪最重要的技术之一。基于无线传感器网络的目标跟踪是当今研究热点之一。本文首先介绍了无线传感器网络的体系结构、节点定位技术和目标跟踪的技术,然后总结了无线传感器网络定位与跟踪需考虑的一些因素包括目标运动状态的建模、常用的运动模型,并且指出传统的目标跟踪算法在无线传感器网络跟踪中面临的问题包括难以确定目标和滤波器的初始状态、间歇性目标丢失以及在杂波与误测量较多的情形下其滤波性能较差。针对上述缺陷,本文基于随机集理论(RFS)和有限集统计(FISST)方法,提出了一种针对无线传感器网络多目标跟踪的粒子PHD滤波器,并且针对传统PHD滤波器的计算复杂度较高的问题,设计了一种分布式粒子PHD滤波器改善其跟踪的实时性能。在基于超声波测距的无线传感网络中,本文首先设计了一种基于HC-SR04阵列的360°超声波传感器硬件节点,而后针对粒子滤波分布式计算的同步需求,提出了一种基于AODV路由协议的时钟同步算法TS-AODV,随后在无线传感器网络时钟同步的基础上,研究并实现了分布式粒子PHD滤波。通过实验证明,粒子PHD滤波算法在传感器网络目标跟踪应用中,能够有效的估计出多目标运动的目标状态、目标数量,并且在杂波较多的情形下性能稳定,分布式粒子PHD滤波算法与集中式粒子PHD滤波算法相比,其单步计算时间较短,从而提高了算法的实时性。
[Abstract]:Target tracking technology has always been a hot topic of exploration. Its application covers military, industrial, commercial, medical and other fields. It has been more than 50 years since target tracking technology emerged. With the rapid development of sensor technology in recent years, the emergence of wireless sensor networks has been highly concerned by all parties, because of its wide coverage, Low cost and diversity of testing methods, Wireless sensor network (WSN) will become one of the most important technologies in the 21 century. Target tracking based on WSN is one of the hot research topics. Firstly, the architecture of WSN is introduced in this paper. Node location technology and target tracking technology, then summarizes some factors that need to be considered in wireless sensor network localization and tracking, including the modeling of moving state of the target, the commonly used motion model, It is pointed out that the traditional target tracking algorithms face some problems in wireless sensor network tracking, including the difficulty in determining the initial state of the target and the filter. The intermittent target loss and the poor filtering performance in the case of more clutter and mismeasurement. In view of the above defects, this paper is based on the stochastic set theory (RFS) and the finite set statistical analysis (fish) method. A particle PHD filter for multi-target tracking in wireless sensor networks is proposed, and the computational complexity of the traditional PHD filter is high. A distributed particle PHD filter is designed to improve its real-time tracking performance. In the wireless sensor network based on ultrasonic ranging, a 360 掳ultrasonic sensor hardware node based on HC-SR04 array is designed in this paper. Then a clock synchronization algorithm TS-AODV based on AODV routing protocol is proposed to meet the synchronization requirement of particle filter distributed computing. Then it is based on clock synchronization in wireless sensor networks. The distributed particle PHD filter is studied and implemented. The experimental results show that the particle PHD filter algorithm can effectively estimate the target state and the number of targets moving in the sensor network. And the performance of distributed particle PHD filter is stable in the case of more clutter. Compared with the centralized particle PHD filter, the single-step computation time of the distributed particle PHD filter algorithm is shorter, thus improving the real-time performance of the algorithm.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9;TN929.5

【参考文献】