Video Event Annotation using Ontologies with Temporal Reasoning

Abstract: Annotation and retrieval tools for multimedia digital libraries have to cope with the complexity of multimedia content. In particular, when dealing with video content, annotation and retrieval tools have to use appropriate knowledge structures that can effectively relate high level concepts to low and mid level visual features and, at the same time, integrate temporal information which is crucial when defining an abstract model for video. In this paper we present a multimedia ontologies that include both linguistic and visual ontology. Moreover provided that appropriate low level descriptors are used to detect simple events, subjects or objects, we propose usage of Semantic Web Rule Language in order to provide a formal definition of complex events based on temporal relations between simple entities. Results for complex event inferencing are shown for the news broadcast domain.

 not available