Ontology-based annotation of images is intended to allow standard and explicit meanings to be assigned to objects and features that an observer perceives there. In practice, interpretation of images in sedimentary petrology dealing with atomic items by themselves in an ontology (such as concepts, attributes and values) is insufficient to capture higher-level semantics (e.g. evidence, findings, justifications). As a response to that problem, the paper proposes two conceptual annotation levels: atomic, and abstracted. We describe here a knowledge-based model that helps users to make their image-interpreting knowledge explicit and apply in interpretation of oil-reservoir rock images. Two concepts are specific to this work: visual chunks, which represent visual patterns of domain terms, and K-graphs, which are used to state knowledge-intensive connections between items of observed evidence and interpretations for this evidence. The essential details of the framework are not at all constrained by the geological application: they can be applied without change in any other subject-areas where reasoning and annotation of images are the primary requirement.
Figures - uploaded by Mara Abel
All figure content in this area was uploaded by Mara Abel
Content may be subject to copyright.