Abnormal Behavior Detection of Vessels Based on Deep Learning Algorithm: Case Study

Abnormal behavior detection for vessels is vital to satisfy the reliability and safety requirements of maritime transportation. The current detection method is manual analysis, which has many defects such as high labor costs, weak stability, poor efficiency, low accuracy, so this passive mode cannot meet the needs of maritime transportation. Using deep learning algorithms to extract valuable information from various types of data and accurately perceive the vessel's behavior is the current research hot spot. This paper firstly preprocessed the data from the automatic identification system (AIS). And then the variable recurrent neural network algorithm was applied to establish the normal model. Next, according to the geographical and seasonal characteristics, several case studies were conducted to show the model's performance. Finally, the research and technological challenges in abnormal behavior detection using deep learning algorithms were summarized.