Road Extraction Using Snakes

This paper presents a methodology for semi-automatic road extraction from digital images using active contour models or snakes. Snakes was proposed almost two decades ago, consisting in a parametric curve controlled by photometric and geometric constraints. This last ones generate internal forces that control the shape of the snakes curve. The photometric constraints generate external forces that attract the snakes curve to the selected feature, i.e., the road. As this snakes-based method is semi-automatic, firstly the operator must describe the road roughly using at least six seed points. From this initial position, the snakes curve moves around at each interation, stabilizing when it coincides with the road. This road extraction methodology was tested with digital aerial images of different resolution. The experimental results obtained with these images showed the method’s potential in accurate and reliable road extraction.