基于传感器网络的目标跟踪应用算法的研究和改进

【作者】 孟沙；

【导师】 罗惠琼；

【作者基本信息】 电子科技大学 ， 计算机系统结构， 2004， 硕士

【摘要】 传感器及其网络的发展使得基于它们的新的应用越来越多地涌现出来,动态目标跟踪就是其中很有前途的应用之一。它对于自然科学里面很多学科的研究,野生动植物研究以及军事情报收集等领域的方法更新及效率改善都具有十分重大的意义。在这些应用中准确地掌握各节点在全局坐标系统中位置的必要性就显得十分突出。通常估算自身位置会引进误差;这类网络中的各传感器之间采用的无线通信方式在进行目标跟踪测量目标位置时会引入各种噪声。为了获得对目标运动轨迹的最接近真实的信息,要求我们采用适当的方法对这些噪声进行过滤。本文首先重点介绍了近年来传感器及其无线网络的发展和应用,介绍无线传感器网络适用的范围和领域,以及设计一个无线传感器网络需要注意的问题和通常的解决措施。还对几种已有的传感器网络的实现方式作了比较。接下来对传感器网络实现其目标跟踪应用的基础平台——传感器网络节点坐标系统的组织生成算法作了介绍。并分析了其可能达到的精度。在此基础上我们对几种可用的过滤算法进行了比较,并因贝叶斯过滤算法的特别适用性而将其引入进来。我们还对贝叶斯算法需要的两个模型——系统模型和测量模型分别针对目标跟踪进行了对应。为了进行简化的近似计算,我们引进了非参数化分布表示方法。在这些理论分析的基础上,我们以先进数学建模仿真平台matlab为依托,建立了共分为七个模块的系统仿真系统,对它们的测试结果进行了单独分析和模块间的关联性分析。最后依据这些模块的测试运行结果,我们以较多的图表来分析说明了贝叶斯过滤器针对目标跟踪这一应用的性能改善的实际效果,获得关于算法优化的信息,证明了在基于多边估计算法的目标跟踪应用中,采用贝叶斯滤波进行噪声过滤对提高整个系统的跟踪精度的效果是十分显著的。更多还原

【Abstract】 With the development of sensors and networks based on them, many new related applications occurred in recent years. In these applications, dynamic target tracking is one of the most hopeful issues. It has significant meaning in method changing and efficiency improvement, which can be applied to the field of subjects in natural science, wild animal study and military information collection. It becomes more important to know node’s position in these situations. The estimation of nodes’ own position may introduce some form of errors in measurement; the wireless communication between sensor nodes is inherently prone to bring noises during position measuring in target tracking task. In order to obtain as possible as nearly true information of target track, it is necessary to filter these noises by some appropriate methods.First in this paper, we put emphasis upon the development of sensor and its network; introduce the proper fields the WSN (Wireless Sensor Network) can be applied to, and the noticeable design issues and their corresponding common solution. Give comparison among several implementation of sensor network. Then give an introduce to algorithm for sensor network to organize a coordinate system, which is the infrastructure for function of target tracking. We analyzed the accuracy it can reach. Based upon this, we compared several filters available, and directed to the Bayesian one due to its prominent applicability. We then give the system and measurement models that Bayesian method defined their respective instance in tracking application. A nonparametric distribution representation is introduced for simplified approximating. According to these theories, we established a simulation system based on the advanced math tool matlab, which can be divided into seven modules. Test results are given both on a per-module and inter-module basis. Finally, we draw a positive conclusion upon the obvious effectiveness that Bayesian filter imposed on target tracking measurement, and this is obtained by a series of graphs and table.更多还原