Path Opening Based Road Extraction for GOKTURK-2 Images

Yesilyurt G., Erturk A.

26th IEEE Signal Processing and Communications Applications Conference (SIU), İzmir, Türkiye, 2 - 05 Mayıs 2018 identifier identifier


Road network detection from satellite images is essential for many applications ranging from city planning to autonomous vehicles. Morphology based approaches have traditionally been commonly used for road extraction applications. Traditional morphological filters provide high performance for structures such as building edges and straight roads. However, real scenes such as satellite images rarely contain straight roads, and hence a more flexible approach that enables to extract curved structures is required. Adjacency graphs based path opening approaches that have been developed recently for this purpose have great potential in the extraction of curved structures and road map extraction from satellite images. In this paper, road extraction from images acquired by the GOKTURK-2 satellite with a path opening based method is proposed. In this work, which is the first road map extraction approach on GOKTURK-2 images, visual and quantitative results on real datasets are presented.