DEVELOPMENT OF ROAD EXTRACTION APPROACH FROM SATELLITE IMAGE

Abstract

There is a growing need for automated procedures that can update linear features from satellite images, such as roads and can be used directly with the existing database of Geographical Information Systems (GIS). Such procedures are in great demand where the geospatial databases have to be updated and maintained regularly. This study aims to propose a methodology for automatic extraction of road network from satellite image. Roads are important components of infrastructure. The accurate information about their location, width and topology are required for planning traffic management, new roads and so on. The model developed using eCognition software has been successfully used for extracting different kind of road networks from IKONOS satellite images. The proposed methodology has been tested on different data and the results are found to be good. However, the connectivity of road network is required to be further developed as the obtained road network has been found to be discontinuous at certain places due to presence of dense trees, traffic flow along the road. It is a challenging problem for further research. The proposed methodology of road extraction, if implemented, can help save manpower and time without compromising on accuracy, and thus can provide an important and reliable input to geospatial databases. These geospatial databases are required for planning transportation networks, road planning projects, urbanization, and so on.