A distance-based trajectory outlier detection method on maritime traffic data

As a result of establishment of Automatic Identification System (AIS) networks, maritime vessel trajectories are becoming increasingly available. Finding outliers in a collection of patterns among AIS trajectories is critical for real time applications ranging from military surveillance to transportation management. In this paper we present a distance-based approach for trajectory outlier detection on trajectory data. First, we apply Density Based Spatial Clustering of Applications with Noise (DBSCAN) to generate patterns from original trajectories. Second, we extract gravity vectors and sample stop points from clusters to retain stop and move information and to reduce the further computation cost. To measure the similarity between trajectory points and clusters, cluster relative distance and cluster angular distance are proposed. Finally, we present experiments on real AIS data at Chinese QiongZhou strait. The experiment results show that our methods can detect distance anomaly and speed/heading anomaly effectively and greatly reduces computation cost.