MARITIME ANOMALY DETECTION USING AIS DATA STREAMS

Marine vessels which are involved in some commercial activities follow certain patterns or paths depending on type of their commercial or business in which they are used. If vessels tend to follow some anomalous behavior, this might indicate that they are involved in some illicit behavior. We proposed a framework for detection of anomalous activities in marine using data streams obtained from Automatic Identification System (AIS). We use recurrent neural network with latent variable modeling and this combined with the AIS messages which embedded. This will create a new representation which will address major issues that will arise when AIS data streams are analyzed and considered as hug set of streaming information, which are unusual time sample and crowded data. This system will significantly concern on trajectory reconstruction, peculiarity and anomaly recognition, identification and vessel type detection.