Additional documentation: digital image processing of QuickBird data for land

Because of the combination of high spatial resolution and low spectral resolution associated with QuickBird imagery it is often difficult to apply traditional classification algorithms (which use spectral properties of individual image pixels) to derive thematic information, particularly in heterogeneous environments. Object-based approaches to classification, which attempt to describe relationships in terms of several categories of object characteristics (in addition to pixel-level information), have generally been shown to yield better classification results in these cases. This document describes the objectoriented approach to land cover classification used in the current study. All analyses were carried out using eCognition v. 4 (Definiens AG, München, Germany).