Omair Chaudhry

Automatic identification of urban settlement boundaries for multiple representation databases

Intuitive and meaningful interpretation of geographical phenomena requires their representation at multiple levels of detail. This is due to the scale dependent nature of their properties. Considerable interest remains in capturing once geographical information at the fine scale, and from this, automatically deriving information at various levels of detail and scale via the process of generalisation. Prior to the cartographic portrayal of that information, model generalisation is required in order to derive higher order phenomena associated with the smaller scales. This paper presents a technique for automatically identifying settlement boundaries based on our understanding of what constitutes ‘citiness’. From this, partonomic structures can be created that link the broad settlement with its constituent parts. The benefits of the resultant system include the automated populating of multiple representation databases (MRDB), better spatial analysis and the creation of semantic reference systems capable of supporting intelligent query or zoom. The creation of such hierarchical partonomic structures provides a very useful framework within which generalisation can take place. The methodology and implementation are presented together with an evaluation of the results. Future developments are proposed.

A functional perspective on map generalisation

In the context of map generalisation, the ambition is to store once and then maintain a very detailed geographic database. Using a mix of modelling and cartographic generalisation techniques, the intention is to derive map products at varying levels of detail – from the fine scale to the highly synoptic. We argue that in modelling this process, it is highly advantageous to take a ‘functional perspective’ on map generalisation – rather than a geometric one. In other words to model the function as it manifests itself in the shapes and patterns of distribution of the phenomena being mapped – whether it be hospitals, airports, or cities. By modelling the functional composition of such features we can create relationships (partonomic, taxonomic and topological) that lend themselves directly to modelling, to analysis and most importantly to the process of generalisation. Borrowing from ideas in robotic vision this paper presents an approach for the automatic identification of functional sites (a collection of topographic features that perform a collective function) and demonstrates their utility in multi-scale representation and generalisation.

Automatic Classification of Retail Spaces from a Large Scale Topographic Database

There is considerable interest in understanding the distribution patterns of different types of retail space, over time, and doing so at a national scale. Yet a lack of suitable data, coupled with poor classification schemas, has stymied efforts to create such a national perspective. This research reports on metrics and classification methodologies that have been applied to large scale topographic data, that afford a systematic classification of certain retail spaces potentially at the national coverage. By analysing the form, composition, extent and patterns of buildings within retail spaces, together with their degree of centrality and levels of access, we demonstrate that it is possible to classify different types of retail space. The research illustrates the utility of fine scale topographic data beyond mere mapping. The article compares three methodologies used for classification (Boolean, fuzzy logic and Bayesian modelling) and evaluates them through comparison with known locations of various retail types as a way of assessing the validity of these approaches. The quality of the results are good, though the work highlights the inconsistency in definitions that currently exist – reflecting, as much as anything, the shifting sands of definitions of various retail spaces that ebb and flow according to consumer needs, and the ambitions of urban planners.

DTM Generalisation: Handling Large Volumes of Data for Multi-Scale Mapping

Abstract title/> Map generalisation is a modelling process in which it is typical that detailed, high dimensional geographic phenomena are reduced down to a set of ‘higher order’, yet more generalized set of phenomena (for example, a large cluster of buildings is reduced to ‘city’). This process of generalisation necessarily requires us to handle large volumes of data which results in high processing overheads. One way of managing this is to partition the data. When geographically partitioning data, we need to partition in such a way that each partition can be generalized without having to consider regions outside any given partition. The focus of this paper is to explore partitioning and generalisation methodologies that can be applied to digital elevation data - the ambition being to derive generalized descriptions of morphology at the National Scale (for the UK). The paper describes and compares two solutions to this problem, and demonstrates how it is possible to apply generalisation algorithms to national coverages.