基于贝叶斯信任模型的无线传感器网络动态事件区域检测算法的研究及其应用

发布时间：2018-05-10 18:53

本文选题：无线传感器网络 + 动态事件区域检测　；参考：《南京邮电大学》2017年硕士论文

【摘要】：无线传感器网络作为一种从现实物理世界获取信息的手段,受到越来越多的重视,其中事件区域检测是无线传感器网络的重要应用之一。但在现实场景中,事件区域通常具有动态性。传统的无线传感器网络模型和算法虽然理论上比较完善,但在现实的应用中表现出诸多的局限。考虑到无线传感器网络自身携带能量有限和易受环境影响的特性,在选择动态事件区域检测算法时要考虑使用复杂度低且容错性较好的检测算法来保证动态事件检测的精度。本文提出了一种贝叶斯模型方法处理无线传感器网络中的动态事件区域检测问题。在本文的模型中,每个传感器节点都与一个虚拟社区和一个信任值相关联,其中信任值是用来定量的衡量其社区中传感器节点的可信度。如果一个传感器节点的信任值小于一个阈值,就代表此传感器节点发生故障,因此在这一时间点,该传感器节点的读数不可信。传感器节点信任值的概念使得本文的模型区分于其他的模型,例如马尔科夫随机场。同时,本文也考虑到了一些实际问题,包括邻近节点的时空相关性,传感器节点中存在故障节点等。根据目标模型,在每个时间步长,每个传感器节点的信任值需要通过粒子滤波算法实时更新。传感器节点是否位于事件区域是由当前传感器节点的信任值决定的。实验结果表明,相比于已存在的动态事件区域检测算法,目标算法在动态事件区域检测问题上具有更好的性能。
[Abstract]:As a means to obtain information from the real physical world, wireless sensor networks (WSN) have been paid more and more attention, among which event area detection is one of the important applications of WSN. However, in real-world scenarios, event areas are usually dynamic. Although the traditional models and algorithms of wireless sensor networks are perfect in theory, they have many limitations in practical applications. Considering the characteristics of wireless sensor networks with limited energy and easy to be affected by environment, the low complexity and fault-tolerant detection algorithm should be considered to ensure the accuracy of dynamic event detection in selecting dynamic event region detection algorithm. This paper presents a Bayesian model approach to dynamic event detection in wireless sensor networks. In this model, each sensor node is associated with a virtual community and a trust value, where the trust value is used to quantitatively measure the credibility of the sensor node in its community. If the trust value of a sensor node is less than a threshold value, the sensor node will fail, so at this time point, the sensor node reading is not credible. The concept of sensor node trust value distinguishes our model from other models such as Markov random field. At the same time, some practical problems are considered, including the temporal and spatial correlation of adjacent nodes, the presence of fault nodes in sensor nodes, and so on. According to the target model, at each time step, the trust value of each sensor node needs to be updated in real time through particle filter algorithm. Whether the sensor node is located in the event area is determined by the trust value of the current sensor node. Experimental results show that the target algorithm has better performance than the existing dynamic event region detection algorithm.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9;TN929.5

【参考文献】