Anomaly Detection in Maritime Ship Trajectory using Deep Learning Approach: A Comprehensive Survey, State-Of-The -Art, and Future Perspective

Prediction of ship trajectories using data from the Automatic Identification System (AIS) has garnered increased attention due to its potential in averting collision incidents and resolving navigational conflicts. Hence, there exists a pressing need to systematically review the literature on deep learning prediction techniques to elucidate their benefits in ensuring maritime safety across various scenarios. This task holds particular importance and relevance in the realm of unmanned vessels coexisting with manned ships, shaping a novel hybrid maritime traffic paradigm in the upcoming era. The present study aims to undertake a thorough review of deep learning methodologies, encompassing Recurrent Neural Networks, Long Short-Term Memory, auto-encoder, and Hybrid methods. The outcomes elucidate the distinctive features of diverse prediction approaches, offering valuable insights for stakeholders to navigate the selection of the most suitable method tailored to specific circumstances. Furthermore, it contributes to identifying the research challenges in ship trajectory prediction and proposing corresponding remedies to steer the course of future investigations. Future perspectives in maritime anomaly detection involve leveraging advanced technologies like Automatic identification Systems (AIS), radars, and Decision Support Tools (DST) to enhance surveillance capabilities and ensure maritime traffic safety. These approaches contribute to improving detection performance, reducing false alarms, and anticipating proper actions in response to anomalous behaviors, ultimately enhancing maritime situational awareness and operational efficiency in complex maritime environments. Anomaly detection in maritime performs a crucial part in enhancing safety of marine traffic and security leveraging advanced technologies and innovative approaches.