We show in this paper that mathematical morphology provides a unified and consistent framework to express different types of spatial relationships and to answer different questions about them, with good properties. We show then how to use these fuzzy relationships in model-based pattern recognition and spatial reasoning under imprecision. Two examples are presented, one where recognition of face features is expressed as non bijective correspondence between graphs representing regions and spatial relations, and one where anatomical expert knowledge involving spatial relationships is used to guide the recognition of brain structures.
展开▼