Semi-automatic road detection in high-resolution satellite images

Road detection from satellite and aerial images is a long-standing research issue for remote sensing community. Digitizing the road network of an area with minimum effort is important not only for interpreting the remote sensing data but also for doing high-level semantic analyses. In this work, a semi-automatic road detection method that makes use of Extended Kalman Filter (EKF) and Daisy feature is proposed for hig-hresolution satellite images. The proposed method starts the road tracking from an operator selected seed point and is able to expand by automatically producing new seed points at road junctions. The main contributions of the proposed method can be summarized as using the road width in the tracking process and using Daisy feature for profile matching to make the matching process more robust than color-based profile matching.