Use this URL to cite or link to this record in EThOS:
Title: Inferring urban land use from very high spatial resolution remotely sensed imagery
Author: Blamire, P. A.
Awarding Body: University of Wales Swansea
Current Institution: Swansea University
Date of Award: 1998
Availability of Full Text:
Access from EThOS:
With the imminent launch of a new generation of very high spatial resolution satellite sensors, 1-5m image data are soon to be available. This thesis explores the potential application of these data to infer information on urban land use. Distinctions are sought between broad categories such as industrial, commercial and residential, as well as finer residential categories indicative of housing age and type. Unlike conventional classification procedures, where land cover is inferred from spectral reflectance, it has been suggested that land use requires information on the spatial and morphological properties of the principal scene objects (e.g. buildings, roads) within the image. However, this has received little formal investigation. This hypothesis is examined quantitatively using Ordnance Survey 1:1250 digital map data (an 'optimum' segmentation of the scene, without the problems of mixed pixels, misclassification, shadowing and occlusion associated with remotely sensed imagery). Differences are observed between areas of contrasting land use, in particular the structure of the road network and the specific composition of buildings (their size and shape) within an area. The ability to extract those scene objects from remotely sensed data is subsequently assessed using airborne imagery, resampled to a number of spatial resolutions between 1 and 10m, and a variety of segmentation procedures (multi-spectral classification, edge-detection and region-growing). This indicates that, while thematic accuracy increases at finer resolutions, it is not possible to extract building and road features as discrete entitles unambiguously and consistently. Further experiments using even higher resolution data (up to 25cm) exhibit similar problems. Despite this, a final set of experiments examine the structural properties of a segmented image. Using a graph-based tool, relations such as the adjacency between regions are shown to vary between areas of contrasting land use. Image structure is also successfully used as a novel means of removing 'noise' from classified imagery.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available