基于时间序列的多维距离聚类异常检测方法

目前轨迹分析的异常行为检测技术以检测位置信息为主,忽略了时空轨迹的轨迹有序性及运动特性,为此提出一种基于时间序列的多维特征聚类异常检测方法,提高广播式自动相关监视(ADS-B)数据异常检测技术的精确性,通过提取ADS-B数据中经度、纬度、速度、航向信息,利用Hausdorff距离计算轨迹数据的多特征相似度,结合层次聚类方法检测轨迹中的异常行为。实验结果表明,该方法能够有效提高飞行轨迹数据的异常行为检测的精确性。

Multi-dimensional distance clustering anomaly detection method based on time series

The abnormal behavior detection technology for the current trajectory analysis mainly detects the position information,ignoring the trajectory order and motion characteristics.A time-based multi-dimensional feature anomaly clustering detection method was proposed to improve the accuracy of the automatic dependent surveillance-broadcast(ADS-B)data anomaly detection technology.By extracting the longitude,latitude,speed and heading information of ADS-B data,the Hausdorff distance was used to calculate the multi-feature similarity of the trajectory data,and the hierarchical clustering method was combined to detect the anomalous behavior in the trajectory.Experimental results show that the proposed method can effectively improve the accuracy of abnormal behavior detection of flight trajectory data.