Abstract: The road is the core element of the urban traffic system,serving as the fundamental infrastructure that connects various regions and carries traffic flow.Overpasses,serving as key structural components within urban road networks,are essential tools for traffic diversion.Therefore,the construction of models for overpass structures is critically important in domains such as urban traffic simulation and navigation positioning.However,traditional overpass modeling methods have issues with weak semantic geometric correlation and insufficient 3D topological representation.To address these issues,this paper proposes a universal segmented and layered overpass modeling method through a deconstruction analysis of overpass structures.This method achieves 3D modeling of complex urban traffic scenes with limited height data and low data costs.The proposed method is capable of modeling overpasses with diverse and complex structures,enhancing the efficiency of overpass modeling and precisely depicting the spatial hierarchy of overpasses.It effectively supports simulation applications such as road navigation and automatic driving,offering inspirational implications for the construction of urban 3D traffic network systems and the simulation of urban virtual environments.

Key words: overpass, traffic scene, segmented and layered, road simulation, autonomous driving