A Symbol Grounding Model for Semantic Interpretation of 2-D Line Charts

Representing symbol grounding is a recurrent subject of discussion in knowledge-based vision systems. The emph{symbol grounding problem} refers to the issue of connecting symbols of a symbol system to their low-level projections within perceptual input. In this paper, we present our ongoing efforts in formalizing an ontologically-founded, domain-independent visual knowledge model that allows one to represent the grounding relation between high-level domain entities and basic image features of 2-D line charts. The characterization of the grounding relations according its ontological status should provide a better criteria in deciding which domain entities can be grounded, improving reusability of our model in multiple interpretation domains and better structuring of the grounding relations. Additionally, we present an algorithm for visual interpretation using the proposed model. The framework of models and algorithm has been implemented in a visual interpretation system for Geology domain. One test case is shown as running example.