Theories of shapes are important for object recognition and for reasoning about the behaviour of objects, both tasks strongly constrained by shape. Whereas the extraction of shape properties has extensively been studied in vision, there is still a lack of qualitative shape descriptions which allow reasoning about shapes with AI techniques in a flexible manner. In this paper we present a qualitative shape description. This description is based on a set of qualitative relations which can be combined to construct arbitrary polygonal shapes. As we are interested in demonstrating how qualitative reasoning approaches can be applied to shape descriptions, our theory is confined to stylised shape representations which are obtainable by applying conventional image processing techniques. We will show how to qualitatively reason about shapes.
展开▼
