ABSTRACT
There is an increasing demand by society for information on the urban environment. One of the most important challenges in this respect is to move from a 2.5D representation of urban area to 3D and 4D - that is, including time. This can only be achieved using remotelysensed images as the primary data source, in conjunction with the analytical techniques offered by Geographical Information Systems (GIS) and knowledge-based approaches. This Chapter presents a snapshot of work under way in this general area, focusing on three separate research projects being carried out at the author's institute, namely (i) change detection in the urban environment, (ii) map updating and (iii) multi-temporal modelling. The studies range in spatial scale. At the small scale, image data from Landsat Thematic Mapper (TM) and the ERS-1 Synthetic aperture Radar (SAR) are merged to help identify urban areas and to distinguish different categories of land cover/land use within them. The second study, which also makes use of data from Landsat TM, employs Delaunay triangulation to delineate the urban-rural boundary. Thi work i based on an initial land cover/land use map generated using a conventional multispectral classification algorithm applied to the TM data. Finally, at the large scale, we report current research to derive vector maps from digitized aerial photography. This is achieved using a Blackboard System, together with an associative memory or a Semantic Network (ERNEST), to extract and interpret the necessary spatial features, and taking into consideration 3D and generalization effects.
