GEOSCAN Search Results: Fastlink


TitleA graph theory approach to road network generalization
AuthorThomson, R C; Richardson, D E
Source17th International Cartographic Conference-10th General Assembly of ICA, Proceedings; vol. 2, 1995 p. 1871-1880
LinksOnline - En ligne
Alt SeriesEarth Sciences Sector, Contribution Series 20041643
Meeting17th International Cartographic Conference-10th General Assembly of ICA; Barcelona; ES; September 3-9, 1995
File formathtml
Subjectsmiscellaneous; remote sensing; topography; vegetation
AbstractThe development of techniques for the automatic integration of remotely sensed data into a GIS environment is one focus of research at Canada Centre for Remote Sensing. One aspect of this effort is the automatic creation of databases of geographic data: which can be reconfigured on demand to yield the information relevant to a given context, e.g. for map compilation. A theoretical framework is provided by a model developed at CCRS for automated spatial and thematic generalization, and techniques are being· developed for automatic structuring, classification and coding of unstructured road network data into a suitable form supporting generalizations. An appropriate classification of road data must take into account features such as surface type and number of lanes, and also functional aspects such as a road's relative importance in linking a given set of locations. Graph representations offer a convenient· means of handling the topological and associated information describing a road network; and the use of graph theory in supporting network analysis and generalization is briefly reviewed. Graph theoretic techniques, such as the shortest path between network nodes and spanning trees, are then shown to provide a solution to the iinportant problem of deriving measures of the functional relevance of network road segments, given a context defined in terms of a set of points of interest. Thus, for a given context, a ·set of rankings reflecting the importance' of the 'segments is- created which can serve as the basis for the attenuation of the network to any required degree for use in map density reduction and generalization. Spanning trees can be used in addition to maintain connectivity between destination points during attenuation. Results from a prototype implementation of the network analysis system are presented. Preliminary tests indicate the effectiveness of the analyses: graph theoretic methods allow·the efficient extraction and handling of the pertinent topological properties 'of a network,. and as such naturally support generalization which aimS'to find the essential or representative characteristics ofa data set t'or a given context.

Date modified: