A Semantic Graph of Traffic Scenes for Intelligent Vehicle Systems

This article presents a semantic graph representation for vision-based intelligent vehicle systems. It can represent the traffic scene with both perceptional meaning of object classes and the spatial relations between them. Using such graphs offers superior performance in terms of both accuracy and robustness. Furthermore, a stereovision-based road boundary estimation system, designed for navigating an intelligent vehicle through challenging traffic scenarios, is introduced, which exemplifies the advantages of the semantic graph.