Road Classification from VHR Imagery

Updating of topographic maps or rapid mapping for supporting crisis management are examples of applications where automated or semi-automated procedures of road extraction are demanded. The paper focuses on development of classification rules for road extraction from very high resolution satellite images and aerial orthoimages. From the methodological point of view, a main emphasis is on object based image analysis. Results of practical tests on QuickBird images from the surroundings of Prague (combination of agriculture, urban and forest areas) are presented. Necessary changes of the classification base for road extraction from aerial images are discussed. A comparison with supervised, per-pixel approach is carried out. Applicability of the developed classification rules for other areas with different land cover are evaluated based on comparison with a national topographic database that was used for accuracy assessment of classification.