RESEARCH ARTICLE Urban Road Network Extraction from Remote Sensing Images Using an Improved F* Algorithm Malika Bendouda 1 • Nasreddine Berrached 1 Received: 27 December 2017 / Accepted: 14 March 2018 Ó Indian Society of Remote Sensing 2018 Abstract A city mapping is essential to various applications such as planning, transport management, vehicle navigation, inter- vention in natural disasters, etc…. For convenience and efficiency, such applications are integrated in a Geographical Information System (GIS) (Bendouda and Berrached, in Etude et re ´alisation d’UREGIS un SIG pour la gestion du re ´seau routier urbain, Magister Thesis, University of Sciences and Technology of Oran Algeria 2009). GIS Map needs real time automatic updating and revisions of the road network databases. However, due to the extreme complexity of the urban environment, there are currently many methods involving the extraction of roads by means of automatic or semi-automatic approaches in rural and sub-urban areas; but in urban environment the majority of these methods failed, due to the complexity of this environment and the complex appearance of the road in the remotely sensed image. In this paper, we introduce a new approach to extract road network in urban area from low resolution satellite images. The proposed method is a modified version of the dynamic programming method for semi-automatic extraction of road network, based on the F* algorithm. The preliminary step is the seeding of points belonging to roads. F* detects the segments that may belong to a road by optimizing certain criteria. Given the complexity of urban areas and the existence of different road categories, we propose an improved version of the classical algorithm F* called PR-F*(Parallel Research-F*). It detects the road segments automatically in many directions. The obtained results are evaluated in terms of quality with respect to completeness and correctness. Keywords GIS Á Urban road network Á Semi-automatic Á Low-resolution Á Dynamic programing Á F* algorithm Introduction Road networks have a great importance in urban planning, map updating, and navigation support as well as effective decision making for military, emergency situations, map- ping traffic management, GPS navigation, public infor- mation service and map updating, etc. (Shi et al. 2014; Wang et al. 2013). In fact, the process of automatic GIS (Geographical Information System) database update needs robust object extraction algorithms to reduce data acqui- sition costs and processing time. Due to the complex appearance of the road in the remotely sensed image, it is difficult to describe different spatial resolution images with fixed parameters or charac- teristics. This explains why the extraction of road infor- mation becomes a very difficult task (Wang et al. 2013). Hence, it is not easy to develop automatically detecting road network from a given image due to several con- straints. For example, the road segments in the image may have different intensity values, with different widths. Moreover, the presence of road junctions and roundabouts may increase the difficulty of the problem (U ¨ nsalan and Sirmacek 2012). Also the roads can be occluded by other objects like buildings, trees and shadow (Lacoste 2004; Herumuti et al. 2013). Especially with the advent of high resolution remote sensing images, even vehicles may par- tially occlude the roads (Peteri 2003). Therefore, advanced & Malika Bendouda vrf06bm@gmail.com Nasreddine Berrached laresi.usto.2015@gmail.com 1 Laboratoire de Recherche en Syste `mes Intelligents ‘LARESI’, De ´partement d’Electronique, Faculte ´ de Ge ´nie Electrique, Universite ´ des Sciences et de la Technologie d’Oran Mohamed Boudiaf, Oran, Algeria 123 Journal of the Indian Society of Remote Sensing https://doi.org/10.1007/s12524-018-0773-3