Abstract
Thematic Mapper (TM) digital data were processed to produce false colour composites (FCCs), ratio images and Colour Ratio Composites (CRCs), Principal Component (PC) false colour composites, and classified images, and these were compared for their geological information content.
Classification by Bayes’ Maximum Likelihood method was investigated, and the best results came from using a summer image and recalculating data to reflectance values. A shadow class was used to improve the quality of the spectral signatures, and it was found that the chosen value of the standard deviation of the band variance was important to the usefulness of the classification.
A comparison of TM and SPOT images (for the two test areas) emphasised the superiority of TM in the spectral domain, and of SPOT in resolution and stereoscopy. The TM FCC of Bands 145 were much more informative than the SPOT FCC, and the SPOT Principal Components did not differentiate rock types as well as TM PCs. However, the superior resolution of SPOT images was apparent at scales of 1:100000 and larger. A combination of SPOT panchromatic and multi-spectral images as a stereopair was found to be a most useful option, combining the advantages of a colour image and a steromodel.