Automatic Road Extraction using Modified Local Vector Pattern

Introduction Extraction of road objects from high resolution imagery has prime significance in road database creation and updation. Moreover, updation of road databases is crucial for many Geographic Information System (GIS) applications like urban planning and change detection. Road extraction from aerial imagery is a challenging task in digital image processing. Extensive research has been done on road detection from aerial imagery. The methods for road extraction can be mainly manual, semi-automatic or automatic. In the manual method, road extraction task is performed manually by human operators (Lu, et al. 2014), (Gheng, et al. 2016). Generally, skilled operators perform this task; however, it is very expensive and requires more time (Shi, et al. 2014), (Z. Miao, et al. 2015), (Moslem and Lepage 2016). Semi-automatic road extraction involves human operators who select the initial seed point(s) for road extraction algorithm. Automatic road extraction does not require any human intervention. Features are extracted automatically (Pandit, Gupta and Rajan 2009). In the recent years, automated extraction of roads has drawn considerable attention due to the requirement for efficient extraction and updation of road data for geodatabases (Revathi and Sharmila 2013), (Bay, et al. 2015). Road networks are essential for precise car navigation, tourism, intelligent transportation systems, traffic simulation, etc. In present times, road network changes at a rapid rate; therefore, periodic updates of road network is essential (Ravanbaksh, Heipke and Pakzad 2007). Automatic road extraction can be classified into three types – 1) knowledgebased, 2) region-based and 3) pixel-based (Coulibaly, et al. 2014), (Keaton and Brokish 2002). Knowledgebased techniques utilise high-level information to identify roads (Saati, Amini and Maboudi 2015), (Sheng, et al. 2013). Region-based techniques classify the imagery into regions to segment the extracted road networks (Han, Bovolo and Bruzzone 2016),

Read Full Article