Ship Type Classification: a Handwriting Signature Verification Approach for Maritime Trajectories

Maritime domain agencies play a critical role in generating situational awareness to detect maritime anomalies and prevent illegal fishing, illegal migration, or even hostile naval activities. Traditionally, these tasks rely especially on human cognitive load, but the use of Artificial Intelligence (AI) can potentially enhance these processes. This paper proposes and analyses several AI-based approaches for vessel type classification based on their trajectory analysis. These approaches draw inspiration from various handwriting signature verification techniques such as deep learning, K-nearest neighbours, support vector machines and convolutional neural networks. Thus, different models were trained using AIS processed data collected from the Black Sea region (Romanian Exclusive Economic Zone). The goal was to obtain a model that classifies non-cooperative vessels when detected only by radar sensors. The trained models aim to identify situations such as illegal fishing, search and rescue operations, or law enforcement activities where vessels may have turned off their AIS transponders. These approaches could have significant implications for enhancing maritime surveillance, particularly in situations where conventional methods are ineffective. Overall, this paper represents a promising step towards improving the safety and security of oceans through advanced vessel type classification techniques.