Ground elevation models and land cover classifers for decimetre resolution urban monitoring

Abstract

This thesis presents methods that enable the generation of quantitative environmental indicators for remotely monitoring urban regions. Its contributions are a new morphological and surface fitting hybrid algorithm for the generation of ground elevation models, a vegetation classifier and significant research into Canonical Variate Analysis with Rational Polynomials (a feature extraction method that normalises the topographic illumination effect). These methods were tested on a 9600 square kilometre, 20cm resolution dataset covering Perth.