基于多维特征终端区航空器异常轨迹识别

Anomalous Trajectory Detection in Terminal Area Based on Multidimensional Trajectory Features

针对航空器轨迹聚类没有充分利用目标的速度、航向等多维特征信息,在发掘轨迹聚类中存在局限性,提出基于多维特征的轨迹聚类方法并基于统计学方法完成异常检测。通过散点相似矩阵确定多维特征,利用多维特征构建多维特征相似矩阵,完成对轨迹的聚类。引入航转角和特征点选择特征轨迹,用多元拟合模型对特征轨迹点拟合,得到航空器特征轨迹表达式,通过计算实验轨迹与位置特征表达式的距离是否大于95%的置信区间距离,完成异常轨迹的检测。在国内某机场用监视数据进行实验,比较结果表明方法具有一定可行性。

For the aircraft trajectory clustering,the multi-dimensional feature information such as speed and head of the target is not fully utilized,and there are limitations in the trajectory clustering.An trajectory clustering method based on multidimensional features was proposed and anomaly detection was performed based on statistical methods.Multidimensional features are determined by scatter similarity matrix,multidimensional feature similarity matrix was constructed by using multidimensional features,clustering of trajectories was completed,and the trajectory of feature angles and feature points were introduced.The feature trajectory points were fitted by multivariate fitting model to obtain aircraft characteristics.The results show that the trajectory expression completes the detection of the abnormal trajectory by calculating whether the distance between the experimental trajectory and the position feature expression is greater than 95% of the confidence interval distance.Experiments with Surveillance data at domestic airport show that the proposed method has certain feasibility.