Abstract

Automated road extraction is an important area of road database creation, refinement and updation. However, the differentiation of road objects for a satellite image is a critical issue. Road extraction from the satellite imagery is classified as semi-automated and fully automated process. Semi-automated extraction requires some human input whereas; the automated extraction process requires no human input. A multistep approach for automatic road extraction is proposed in this paper. Proposed algorithm is implemented for different high resolution images. Experimental results show that the algorithm extracts the main roads from satellite images with about 91% accuracy. However, it is less accurate to detect secondary roads like lanes and by-lanes.