A Pattern Recognition Analysis of Vessel Trajectories

The automatic identification system (AIS) facilitates the monitoring of ship movements and provides essential input parameters for traffic safety. Previous studies have employed AIS data to detect behavioral anomalies and classify vessel types using supervised and unsupervised algorithms, including deep learning techniques. The approach proposed in this work focuses on the recognition of vessel types through the Take One Class at a Time (TOCAT) classification strategy. This approach pivots on a collection of adaptive models rather than a single intricate algorithm. Using radar data, these models are trained by taking into account aspects such as identifiers, position, velocity, and heading. However, it purposefully excludes positional data to counteract the inconsistencies stemming from route variations and irregular sampling frequencies. Using the given data, we achieved a mean accuracy of 83% on a 6-class classification task.

Automated summary

This paper proposes a vessel type recognition method based on the TOCAT classification strategy, which uses a collection of adaptive models trained with radar data. The models consider factors such as identifiers, velocity, and heading to achieve an average accuracy of 83% on a 6-class classification task.