Road network extraction from remote sensing using region-based mathematical morphology

In this paper, we introduce an efficient and automatic method for road extraction from satellite or aerial images. It builds upon an existing work based on (incomplete) path opening/closing, morphological filters able to deal with curvilinear structures. We propose here to apply such techniques not on pixels directly but rather on regions representing road segments, to improve both efficiency and robustness. To do so, we map road segments by rectangular areas, which are identified in a first step. We then adapt the morphological filters to process regions instead of pixels. The resulting path filter allows for connection of road segments, in order to identify roads of minimal length. Robustness to occlusion is ensured through the adaptation of the incomplete strategy to our context, while better discrimination between road segments and other objects relies on background knowledge through an hit-or-miss transform. Preliminary results obtained on several satellite images are promising.