A Logical Approach to Incorporating Qualitative Spatial Reasoning into GIS (Extended Abstract)

The paper explores the application to GIS of formal logical representations and reasoning algorithms for manipulating qualitative spatial information. We consider a number of different formal representations from the point of view of expressive power, 'naturalness' and computational tractability. We find that, whilst there are tradeoffs between these properties, it it is possible to compute effectively with a quite expressive set of spatial relations. Specifically by using an encoding into intuitionistic propositional logic (Bennett. 1994) it is possible to construct a decision procedure for a large vocabulary of topological relations, which runs in polynomial time (Nebel 1995). A considerably more expressive and arguably more natural representation is provided by the lst-order Region Connection Calculus (RCC) (Randell, Cui and Cohn 1992). The basic theory contains just one primitive, the binary connection relation, C(x, y) but by adding a convex-hull operator, cony(x) (whose value is the smallest convex region of which x is a part), a more expressive extended language is obtained. We investigate the application of RCC to characterising geographical features, such as those illustrated in figure 1. We explain how the notion of a 'bay' can be straightforwardly described in terms of convex hulls; and features of rivers such as 'meanders', 'ox-bow loops' and 'ox-bow lakes' can also be characterised.