基于椭球模型的无线传感器网络的局部离群点检测

摘要

针对现有的无线传感器网络(WSNs)的局部离群点检测算法由于存在未考虑监测环境的异质性而造成邻域划分不准确、检测精度低的问题,提出适用于异质监测环境的基于椭球模型的无线传感器网络的局部离群点检测算法.算法用椭球模型刻画数据分布,节点间只传输模型参数,用椭球参数式方程计算椭球间的相异度；将数据分布的不一致性引入到邻域划分的过程中,最终利用传感数据的时空关联性来确定局部离群点.实验结果表明,提出的算法具有通信量低、检测精度高和误检率低的优点.%The existing local outlier detection algorithms in WSNs have some drawbacks, such as the inaccurate of neighborhood and low detection precision due to the ignorance of the heterogeneity of monitoring environment. Therefore, this paper proposed a local outlier detection algorithm for WSNs based on ellipsoids for the non-homogeneous environments. This algorithm characterized the distribution of data at each sensor by ellipsoids. It transported the parameter of ellipsoids among nodes and adopted the method of ellipsoid parameter equation to calculate the dissimilarity between the ellipsoids, and determined the neighborhood of the node considering the inconsistency of data distribution. Finally, it took the temporal and spatial correlation of sensing data to detect local outliers. The experimental results show that this algorithm can achieve great accuracy of detection rate, low false alarm rate and low communications volume in non-homogeneous environment.