Relation-based similarity

Abstract

Similarity queries constitute an active area in spatial query processing. The paper addresses the problem of qualitative similarity based on spntial relntions. On the one hand spatial similarity entails mechanisms for representing and reasoning on spatial relations, while on the other it introduces n high level of uncertainty. Several spatial queries can be rather fuzzy and user or application dependent. Moreover, relations such as near, norhzust etc. lack universally accepted semantics, and ns a result their processing in Spatial Datnbnses and GIS has to provide a high level of flexibility in order to satisfy renl-life needs. Our work extends the notion of conceptual neighbourhood (originnlly defined for ID spnce) to include higher dimensions and proposes n unified multiresolution framework for the hnndling of topological, directional and distance relations. We discuss how object and image similarity queries can be effectively handled and show how uncertninty con be seamlessly incorporated in our model.