Accurate semantic modelling of urban road infrastructure is critical for digital twins, traffic simulations, and smart city planning. This study presents a structured methodology to transform road elements segmented from urban point clouds into CityGML 3.0-compliant representations. Leveraging CityGML’s hierarchical Transportation module, the approach introduces a multi-level granularity framework—<em>area</em>, <em>way</em>, and <em>lane</em>—for representing road components like sidewalks, driving lanes, and parking areas. Following geometric pre-processing, segmented surfaces are semantically mapped into appropriate CityGML classes using a rule-based mapping strategy, enriched with descriptive attributes and hierarchical identifiers. The resulting XML-based datasets were validated and visualized using industry-standard tools such as FME, QGIS, and 3DCityDB, demonstrating successful integration into city-scale digital environments.