A hybrid artificial intelligence system for securing a maritime zone based on historical and real-time data analysis

Securing an area where a maritime event (e.g., a sailing race) occurs is essential to ensure the smooth running of such event. We discuss and detect the risk which may be caused by a ship passing close to a protected area. Based on information provided by fused data coming from a radar, an AIS and cameras, a rule-based classifier has been developed. It aims to analyse real-time data like position, speed and course in order to detect potential threats beforehand. The expert system developed here is built on a two-stage event detection where threats rules are aggregated from simple event detection. In addition, a historical maritime traffic data set is used to train a simple but efficient machine learning model based on hexagonal hierarchical geospatial indexing system H3.