A framework for anomaly detection in maritime trajectory behavior. Issue 1 (April 2016)

• Record Type: Journal Article
• Authors: Lei, Po-Ruey
• Is Part Of: Knowledge and information systems. Volume 47:Issue 1(2016:Apr.)
• Page Start: 189
• Page End: 214
• Publication Date: 2016-04
• Subjects: Maritime trajectory -- Trajectory data mining -- Anomaly detection; Expert systems (Computer science) -- Periodicals; Information storage and retrieval systems -- Periodicals; 006.33
• DOI: 10.1007/s10115-015-0845-4
• Languages: English
• ISSNs: 0219-1377

Abstract

Abstract Rapid growth in location data acquisition techniques has led to a proliferation of trajectory data related to moving objects. This large body of data has expanded the scope for trajectory research and made it applicable to a more diverse range of fields. However, data uncertainty, which is naturally inherent in the trajectory data, brings the challenge in trajectory data mining and affects the quality of the results. Specifically, unlike trajectory collected from vehicles moving along road networks, trajectory data generated by vessels moving free in maritime space have increased the difficulty of sea traffic analysis and anomalous behavior detection. Furthermore, due to the huge volume and complexity of maritime trajectory data, it is hard to define the abnormality of movement behavior and detect anomalies. Additionally, traditional analysis and evaluation by human intelligence is overloaded with the dramatic increasing in amount of maritime trajectory data and is an inefficient approach. Thus, an effective automated method for mining trajectory data and detecting anomalies would be a valuable contribution to maritime surveillance. This paper explores the maritime trajectory data for anomalous behavior detection. We propose a framework for maritime trajectory modeling and anomaly detection, called MT-MAD. Our model takes into account the fact that anomalous behavior manifests in unusual location points and sub-trajectories in the spatial domain as well as in the sequence and manner in which these locations and sub-trajectories occur. This study therefore began by identifying outlying features required for anomaly detection, including spatial, sequential, and behavioral features. We then explore the movement behavior from historical trajectories and build a maritime trajectory model for anomaly detection. The proposed model accurately describes movement behavior and captures outlying features in trajectory data. We then developed an anomaly detection algorithm based on this model in which an indicator is used to evaluate suspicious behavior and scores trajectory behavior according to the defined outlying features. Experiment results demonstrate that the proposed MT-MAD framework is capable of effectively detecting anomalies in maritime trajectories.