Ontologies for probabilistic situation assessment in the maritime domain

Abstract

In the maritime domain, surveillance systems are used to track vessels in a certain area of interest. The resulting vessel tracks are then displayed in a dynamic map. However, the interpretation of the dynamic environment, i.e., the situation assessment (SA) process, is still done by human experts. Several methods exist that can be used for automatic SA, but often they are based on machine learning algorithms and do not include the knowledge of the decision maker. In this article, we describe how expert knowledge can be used to determine models for automatic SA. The knowledge about situations of interest is modeled as an ontology, which can be transformed into a dynamic Bayesian network (DBN). The main challenge of this transformation is the determination of the structure and the parameter settings of the DBN. The resulting DBN can be connected to real-time vessel tracks and is able to estimate the existence of the situation of interest in every time step.