Use this URL to cite or link to this record in EThOS:
Title: The use of spectral and spatial information in the classification of aircraft, rocket and satellite imagery
Author: El-Laham, Nabil Mohamed Aly
ISNI:       0000 0001 3443 1747
Awarding Body: University of London
Current Institution: Royal Holloway, University of London
Date of Award: 1976
Availability of Full Text:
Access from EThOS:
Access from Institution:
The research described in the thesis is concerned with the analysis of imagery obtained from aircraft, rockets, and manned and unmanned spacecraft. The imagery is of two types, one being conventional photographic imagery using the visible and/or the near infrared parts of the spectrum. The other is derived from multispectral scanners carried by Landsat 1 and 2 spacecraft in which information is produced separately for each of the two visible and two infrared spectral bands. The analysis is directed towards the automatic machine classification of the images for different applications and also incorporates a comparison of different methods of classification. One method of analysis is based solely upon the spatial variations of intensity in a single waveband and provides textural information for a scene. This variation has been examined using both one and two-dimensional Fourier analysis to provide a mathematical description of the textural changes. Use of spectral information allows greater variation in the methods of analysis which fall into two principal classes. In the unsupervised method group classification of the images is made according to some similarity measure between the spectral content of each picture element and with no a priori information on their identities. The relative advantages of this method as applied to the imagery considered here have been examined in relation to those possessed by the alternative approach, using the supervised method ofanalysis, which requires a priori information on the type and degree of classification which is required. The supervised method employed here assumes a Gaussian distribution for the spectral intensities of each class and uses a maximum likelihood decision rule for the classification. The extent to which the characteristic spectral signature of each class can be spatially extrapolated and still yield acceptable results has been examined, as has the effort of incorporating a 'threshold' below which classification does not take place.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Computer Science