Road Extraction from a High-Resolution Satellite Image

Abstract

An efficient and general software for getting the idea of the road network in a given satellite image (normally a .tiff file) is to be developed. Satellite images are normally in .tiff file format which is most suitable where the information in the image is of more concerned. In color-based classification, the intensity values are specified for a particular set of objects. The image is then converted to bilevel format with the specified objects being highlighted. The satellite images are in RGB format. Existing algorithms suffer from defects of being slow and memory intensive. Therefore the main objective is to develop efficient software for extracting the road network in a given image automatically & semi-automatically with minimized user interaction. This dissertation presents algorithms for extracting the road network from satellite images in a platform independent language.