Anomaly Detection of Spatiotemporal Sub-trajectories Based on Local Outlier Factor

In the field of early warning and surveillance, it is of great significance to detect abnormal behaviors of targets from massive spatiotemporal trajectories for situation cognition. At present, most trajectory anomaly detection methods are based on the whole trajectory and are insensitive to local anomaly sub- trajectories. In this paper, Spatiotemporal Sub-trajectories Anomaly Detector (SSAD) is proposed. Firstly, Hausdorff distance measurement is applied to sub-trajectory in spatiotemporal dimension to solve the problem of insensitivity to course and velocity anomalies. Then, Spatiotemporal Sub-trajectories Hausdorff Local Outlier Factor (SSHLOF) is constructed based on Local outlier factor (LOF), which can effectively identify local abnormal behaviors in the whole trajectory. Finally, SSAD is implemented in simulation and measured trajectory data respectively. Experiment results show that our method can effectively detect local abnormal behavior of sub-trajectory, which has a good application prospect in intelligent surveillance tasks.