Automatic road extraction from aerial images plays a crucial role in transportation
applications. It provides means for creating and updating transportation databases,
which are needed for traffic management and automated navigation. Up-to-date representation
of transportation networks is essential for critical decision making, such as emergency
response and incident management. Road networks extracted from aerial or satellite
images, will generally be incomplete and fragmented. Therefore, the extracted road
segments have to be connected to form a complete road network. Fragmentation occurs
due to shadows cast onto the road by other objects, such as buildings or trees,
or false road hypotheses such as roofs of houses. The ESP algorithm is able to extract
vessel segments which are very similar to the road segments.