Extracting Centerlines from Dual-Line Roads Using Superpixel Segmentation

Abstract

Extracting centerlines from dual-line roads is very important in urban spatial analysis and infrastructure planning. In recent decades, numerous algorithms for road centerline extraction based on the vector data have been proposed by various scholars. However, with the continual development of computer vision technology, advances in the corresponding theories and methods, such as superpixel segmentation, have provided new opportunities and challenges for road centerline extraction. In this paper, we propose a new algorithm called superpixel centerline extraction (SUCE) for dual-line roads based on the raster data. In this method, dual-line roads are first segmented using a superpixel algorithm called simple linear iterative clustering. Then, the superpixels located at road intersections are merged to generate connection points from their skeletons. Finally, the centerlines of roads are generated by connecting the center points and edge midpoints of each superpixel. To test the proposed SUCE method, the vector data of roads at a scale of 1:50 000 from Shenzhen, China, and the raster data of roads at the 18th level from the Tiandi map are used. Compared with a traditional method in ArcGIS software (version 10.2) based on the vector data and four existing thinning algorithms based on the raster data, the results indicate that the proposed SUCE method can effectively extract centerlines from dual-line roads and restore the original road intersections while avoiding burrs and noises, both for simple and complex road intersections.