How Useful Is Formal Knowledge Representation for Image Interpretation?

In this position paper we provide arguments for the following main points: (1) Formal knowledge representation and well-founded inference processes can help to clarify, unify and support the construction of image interpretation systems. (2) Today's terminological systems (like LOOM) fail to support basic image interpretation inference strategies, in particular hypothesis generation and spatial reasoning. (3) Terminological systems can be extended to provide inference services useful for image interpretation tasks. 1. Introduction The topic of this contribution is formal knowledge representation and its use for image interpretation. We do not raise a discussion about any kind of knowledge representation - there can be no doubt about the need of explicit knowledge representation in knowledge-heavy interpretation tasks. Our point of discussion is the usefulness of representing knowledge in a formally accountable way, typically based on some kind of logics. In particular we look at termin..