On The Assessment of Generalisation Consistency Vasilis Delis Computer Technology Institute Kolokotroni 3, 26221, Patras, Greece Computer Engineering and Informatics Department, University of Patras delis@cti.gr Thanasis Hadzilacos Computer Technology Institute Kolokotroni 3, 26221, Patras, Greece thh@cti.gr Abstract Theory, algorithms, techniques and tools for producing a generalisation of a map have long been available. In this paper we study the inverse problem, namely, given two maps L and M, whether there exists a generalisation G, such that L=G(M). Answering this problem can help with fundamental issues of consistency in multiresolution databases. We view such a database as a collection of map layers depicting the same geographic area at different levels of detail, related through a generalisation hierarchy. From an engineering perspective, multiple representations, of which multiresolution maps is a special case, imply redundancy, which is a threat to the integrity of a database. For integrity control we need a set of tools that ensure that the metric and topological properties of a map layer are retained or monotonically decreased along the generalisation hierarchy. In this paper we study the former, i.e. we propose a framework for the assessment of metric consistency between two map layers. 1 Introduction In a way, cartography is the art of generalisation (when abstracting the real world to create a map or deriving a small scale map from a large scale one) which has long been a difficult and highly subjective problem. "It is the process of reducing the amount of detail in a map so that the character or essence of the original features is retained at successively smaller scales. In the context of topographic map generalisation this usually involves omission, aggregation, simplification, displacement, exaggeration and symbolization of either individual features or groups of features" [19]. Intuitively it involves a transition between two different perception levels. In undertaking this process, a cartographer makes use of his knowledge on the source representation, the intended purpose of the target representation as well as his knowledge of cartographic conventions and the real world [16]. The rules for such transformations are extremely context dependent: relative size, relative scarcity of the entity, “importance” (for the purpose of the specific generalisation). A considerable body of literature exists on map generalisation (see [3, 15] for general treatments). Modern tools for digital generalisation, designed and implemented to exploit the latest in computer science, like object orientation [13], expert systems [19], neural networks [23], or case-based reasoning [12], clearly show