Skip to Main Content
This paper presents a technique for urban road network extraction from high-resolution multispectral satellite imagery. The imagery is first classified using a pixel-based fuzzy classifier and the urban land cover classification are then further refined using an object-based classification approach. The road network extraction technique iteratively identifies line segments in the urban land cover classification and then grows these line segments in the urban land cover classification and then grows these line segments to track roads through occluded areas and around corners. This result of this technique is compared to the road network obtained by calculating the morphological skeleton of the classification image and found to have a significant increase in correctness, however there is a decrease in the completeness measure.