On the representation and exploitation of context knowledge in a harbor surveillance scenario

Maritime surveillance involves gathering and integrating a large amount of heterogeneous information of variable quality to provide diverse decision makers with reliable knowledge about situations and threats. This requires information processing at all fusion levels while taking into account contextual information. Context is especially important for harbor surveillance, one of the most challenging maritime scenarios due to the high number of different vessel types, the coexistence of very diverse operations, the multiple agencies and countries involved, etc. Successful processing of both contextual and transient observed information requires a reusable representation of the harbor domain, as well as effective reasoning methods. This paper discusses an approach to designing a hybrid harbor surveillance system combining ontology-based context representation, deductive reasoning for detection of abnormal objects from their characteristics and behavior, and abductive reasoning under uncertainty.