Road Extraction from Remote Sensing Imagery Based on Road Tracking and Ribbon Snake

In this paper, a semi-automatic road extraction algorithm is proposed. The algorithm starts from a Euclidean distance transform to convert the original remote sensing image into a distance image, which presents all roads in the darker color and can be used as the input image for further process. Then, the contrast of the distance image is enhanced and the skeletons of roads are found. Following these steps, road tracking based on template matching is implemented in order to integrate the discontinuous segments of skeletons. The tracker searches the areas along the direction of the road to extent the skeleton, which produces longer initial curve for snake models. Furthermore, the geometric features of the road are utilized to construct a region-based constrain energy to enforce the contrast maximizing. The experimental results show that the improved ribbon snake extracts the interested roads rapidly and accurately.