基于轨迹信息熵分布的异常轨迹检测方法

针对异常轨迹检测多特征检测和检测单元造成的检测效率低等问题。提出一种基于轨迹信息熵分布的异常轨迹检测方法。该算法根据轨迹偏转角与速度将轨迹分割成若干轨迹段，计算轨迹段间加权多特征距离判断轨迹间相似度，进而完成轨迹聚类并计算出每类代表性轨迹，然后对待检测轨迹进行分割，利用代表性轨迹计算每个轨迹段的信息熵，通过比较轨迹信息熵大小及其分布特点实现异常轨迹检测。大西洋飓风数据仿真实验结果表明该方法提高了聚类效果，克服以整条轨迹检测效率低的缺点，提升了异常轨迹检测算法的有效性。

Trajectory outlier detection based on trajectory information entropy distribution

In view of fact that The detection efficiency of the multi-feature detection and detection unit for trajectory outlier is inefficient, this paper proposed a new method named TOD-TIED(Trajectory outlier Detection Based on Trajectory Information Entropy Distribution). Firstly, the algorithm partitioned a trajectory into a set of trajectory segments according to corner and velocity, then calculated the weighted multi-feature distance between to determine the similarity between trajectory segments. Finally,grouped trajectories into clusters and calculated representative trajectory. the algorithms partitioned the trajectory into a set of trajectory segments, then the information entropy of each trajectory is calculated by using the representative trajectory, finally the abnormal trajectory detection is detected according to the trajectory information entropy and its distribution characteristic. The simulation results of Atlantic hurricane data show that this method can improve the clustering effect and overcome the shortcomings of inefficient detection with the whole trajectory, and improves the effectiveness of the outlier trajectory detection algorithm.