Authors:

Kostas Patroumpas, Alexander Artikis, Nikos Katzouris, Marios Vodas, Yannis Theodoridis, Nikos Pelekis

Abstract:

We present a system that combines intelligent online tracking with complex event recognition against streaming positions relayed from numerous vessels. Given the vital importance of maritime safety to the environment, the economy, and in national security, our system leverages the real-time acquisition of vessel activity with geographical and other static information. Thus, it can offer timely notification in emergency situations, such as intrusion into marine preservation areas, loitering, and unsafe sailing. Thanks to a mobility tracking module, evolving trajectories generated by massive positional updates can be compressed online into concise, but reliable synopses per ship, retaining only salient motion features within a sliding window. These features are exploited by a complex event recognition module that detects suspicious situations of interest to maritime authorities. We conducted a comprehensive empirical validation against a real dataset of traces collected from thousands of vessels. Our results confirm the scalability and approximation accuracy of the proposed system, and thus demonstrate its potential for effective, real-time maritime monitoring

Link:

Event Recognition for Maritime Surveillance