Victor Mesev

---

Abstract | Full Paper (PDF, 3.7 MB)

The notorious difficulties of sensing urban areas remotely have traditionally limited accuracy and thematic detail to coarse generalizations. With the advent of very high spatial resolution satellite imagery comes the opportunity to generate much finer scale inventories of the geography of cities. However, urban reconstruction from imagery is still limited to the classification of physical land covers; and land use still, on the whole, is inferred from the spectral composition of land covers. What is needed to improve urban image pattern recognition and classification accuracy is for GIS to keep pace with the demand for finer spatial resolution ancillary data at higher temporal frequency and reliable spatial consistency.

In Great Britain, the national mapping agency, the Ordnance Survey, is engaged in a rolling program of high spatial resolution digital records of every postal delivery point. The database is known as ADDRESS-POINT® and is precise to within 0.1m of the actual delivery property. In this paper, on-going research will introduce mailing addresses from ADDRESS-POINT as a disaggregate alternative to census-based zonal units, where, although lacking depth in socio-economic information, the address points nevertheless contain valuable spatial attributes representing the spacing of residential and commercial units. Spatial measurements of the density and linear arrangement of groups of address points (using nearest neighbor indices) allow unique characterizations of, and facilitate considerable differences between, types of residential and types of commercial developments. Such characterizations and differences are then explored within an image pattern recognition scenario as a means of inferring land use from land cover classified from high spatial resolution remotely sensed data.

Victor Mesev
School of Environmental Sciences
University of Ulster
Coleraine, BT52 1SA, Northern Ireland
Email: tv.mesev@ulst.ac.uk

back | top