Spatio-temporal trajectory anomaly detection based on common sub-sequence

With the rapid development of GPS positioning and wireless communication, more and more trajectories are collected. How to accurately and efficiently detect abnormal trajectories from a large number of trajectories has become a focused issue. The similarity measurement method adopted by the existing abnormal trajectory detection technology often ignores the situation that the abnormal sub-trajectory has enough neighbors. If a trajectory is composed of multiple such sub-trajectories, this anomaly will not be detected. At present, the trajectory outlier detection algorithm based on common slices sub-sequence(TODCSS) has improved the above problems. However, it is not accurate enough in feature extraction. Its detection scope is limited to 2D-plane and the time dimension is ignored, so it can’t detect abnormal vehicle behaviors such as multiple stops, detention, too slow speed and so on. Based on the above problems, this paper proposes a spatio-temporal trajectory anomaly detection based on common sub-sequence (STADCS). Firstly, in order to obtain accurate and reasonable similar trajectories, the length of sub-trajectory is added to the common sequence of trajectories, and non-common parts between two trajectories are added to the similarity measurement. Then the time is added to detect trajectories of time anomalies. It improves the accuracy and rationality of detection. Finally, we conducted experiments on real datasets and used F1 − measure to evaluate the accuracy of this algorithm. Compared with existing algorithms, the accuracy of STADCS is improved by about 15.15%.