High spatial resolution satellite imagery has been widely used in mapping, environmental monitoring, disaster
management, city planning, because of its favorable visual effects, plentiful texture information, accurate positioning etc.
Traditional classification methods which face to the medium/low-resolution satellite data have been proved not fit for the
high resolution image processing. The object-oriented classification method can resistance the salt and pepper effect,
because it based on patches of spectrally similar pixels which have been produced by image segmentation. In this paper,
a hierarchical framework that based on the stratified classification idea is proposed and applied to the land cover
mapping of city. This stratified framework integrates the object-oriented multi-scale segmentation technology and
quantification of image object features. The scale parameter of segmentation is the key factor during the framework
building. In the study, Scottsdale, Arizona state, USA, is selected as the study area because of its plentiful spatial features
and beautiful sight. The overall accuracy of the land cover classification is 82.58%, the Kappa Coefficient is 0.80 and the
user's accuracies of the most land-objects are exceeding 85%. The study is demonstrated using the object-oriented image
analysis software, Definiens Developer 7.0, which can be integrated with other spatial data in vector-based geographical
information system (GIS) environments.