I. Introduction
RAPID update of the inventory of urban road features, on a regular basis, is of great importance for transportation infrastructure management departments, as well as for intelligent transportation-related applications, including driver assistance and safety warning systems [1], [2] and autonomous driving [3], [4]. Accurate and real-time information regarding current road conditions, traffic flow, and the surrounding environment forms significant inputs to the Intelligent Transportation Systems. The absence, or lack of visibility of necessary traffic signals, is often the cause of terrible traffic accidents. Thus, effective advance detecting and monitoring of potential disasters on the road can reduce casualties and improve traffic safety. According to manuals in different countries, documentation of urban road feature inventory includes road geometries (e.g., longitudinal and transverse slopes, road curvatures, lane width, number, and direction of travel), road surface features (e.g., road markings, manholes, sewer wells, and cracks), and roadside infrastructure (e.g., light poles, traffic signposts, power lines, and bus stations). Not only can the vectorized data be considered by transportation agencies to maintain, repair, and reconstruct current road signals, but they also provide auxiliary information for intelligent vehicle applications to make decisions and improve driving safety. Thus, effective, automated extraction of urban road facilities, such as street light poles, traffic signposts, and bus stations, can assist in rapid update of urban road feature inventories.