Abstract. In this paper, a novel architecture for high-level scene interpretation is introduced, which is based on the generation of rules from an OWL-DL ontology. It is shown that the object-centered structure of the ontology can be transformed into a rule-based system in a native and systematic way. Furthermore the integration of constraints - which are essential for scene interpretation - is demonstrated with a temporal constraint net, and it is shown how parallel computing of alternatives can be realised. First results are given using examples of airport activities.