I. Introduction
Road information plays the fundamental role in many applications, such as smart city construction, urban planning, vehicle navigation, and automatic drive [1], [2], [3]. To obtain up-to-date road information in large-scale areas, extracting road network from high-resolution remote-sensing images has become a promising method. High-resolution remote sensing imagery has become the main data source for extracting road regions and updating geospatial databases in real time [4]. Although road extraction from high-resolution satellite imagery recently gained considerable attention, this task remains challenging owing to irregular and complex road sections and structures [5]. Therefore, various methods for extracting roads from high-resolution satellite imagery have been proposed.