Logical handling of uncertain, ontology-based, spatial information

Abstract

The paper describes a logical framework for handling uncertain spatial information, and merging it when it comes from multiple sources. For this purpose, we use a simple logical formalization for spatial ontologies and for property ontologies relative to different universes of discourse (these ontologies only involve subsumption and mutual exclusiveness relations), since spatial information typically associates properties to sets of parcels that are themselves described in terms of the spatial and/or property vocabularies appearing in the ontologies. Apart from the ontological information describing relations between vocabulary labels, we propose to represent a piece of spatial information as a pair, called “attributive formula”, associating a property formula to a set of parcels (represented by a spatial formula). A set of inference rules is given in order to be able to reason from these attributive pairs. Then, we examine how uncertainty can be encoded in attributive information, using possibilistic logic in a reified manner with respect to parcels. Another important issue pointed out in this paper is that there are two ways to link a property to an area. A first meaning is that the property is true everywhere in the area, a second meaning is that the property is at least true somewhere in the area. This distinction is necessary in order to be able to use both ontological information (which can be encoded by “everywhere” attributive-formulas) and attributive information (which contain the two kinds of attributive-formulas). Lastly, the paper studies how information fusion problems can be handled in the context of spatial data. The problems encountered do not come only from the uncertainty and the possible inconsistency of information as in any information fusion situations, but also from the fact that sources may use different space partitions and may not explicitly specify the somewhere or everywhere reading associated to the information.