A Comprehensive Review of Multimodal Artificial Intelligence Techniques in Maritime Surveillance Using AIS, SAR, and Optical Data Fusion

Maritime surveillance lies at the centre of the maintenance of world trade security, environmental protection, and national security. With the rapid exponential growth of shipping traffic and the constant danger of criminal or "dark" shipping activities, the demand for multimodal data-intensive systems that can integrate auxiliary information from heterogenous sensors such as Synthetic Aperture Radar (SAR), Automatic Identification System (AIS), and optical imaging has grown substantially. This review gives an integration of recent advancements in machine learning (ML) and deep learning (DL) techniques for sea monitoring that values multimodal data fusion and explainable AI (XAI) protocols. A bibliometric evaluation of 2016-2025 publications demonstrates a growing transition from single modality to fusion and combination of systems, where the integration of SAR and AIS has become the most promising paradigm for real-time vessel detection and anomaly identification. The review methodologically surveys performance across datasets, fusion methods, failure modes, learning processes, and dynamic patterns in the current developments in the fields of radar-based, optical-based, and multimodal methods, research issues like data heterogeneity, sparsity of labelled data, scalability, and the lack of interpretability for deep models. By integrating cross-disciplinary understanding, the present review also serves to inform future research towards sound, explainable, and operationally feasible multimodal maritime surveillance systems.