Geometric refinement of road networks using network snakes and SAR images

In this paper, a new approach for the geometric refinement of road networks using network snakes and SAR images is presented. Network snakes are based on the well-known active contour models, but in addition to the image energy and internal energy the topology is introduced into the optimization process. This graph-based active contour method enables a complete topological and shape control during the object delineation. The method is applied to the geometric refinement of road networks to improve and correct GIS-databases as a basis for traffic navigation or infrastructure planning purposes. The proposed approach is either able to deal with roads from a database as initialization in an automatic system or, alternatively, within an interactive framework to derive a geometrically optimized road network. The derived results using SAR images are evaluated with reference date to demonstrate the benefit and transferability of network snakes.