Generating photorealistic urban digital twins with Occlusion-Aware texture mapping

• Automated texture mapping for CityJSON 3D city models. • Ray-tracing method for optimal facade image selection. • City-scale, occlusion-aware texture assignment approach. • File and texture size reduction via optimization methods. • Blender add-on for viewing textured CityJSON models. High resolution textures play a vital role in enhancing realism of urban digital twins, particularly in applications such as urban planning, energy analysis, and geospatial visualization. While textures can provide valuable surface-level details—such as material types (e.g., windows, green façades, bricks) and sunlight reflectance (albedo), these details are often missing in existing 3D city models. Existing 3D city models lack very high-resolution textures, limiting their realism and use cases. Manual texture mapping is labor-intensive and becomes impractical at city scale, often requiring days to weeks of effort. To address this gap, we present the CityJSON Texture Mapper, a Python-based software tool for the automated texture mapping of CityJSON 3D city models from oblique and nadir aerial imagery. The tool assumes that interior and exterior orientation parameters are available, ensuring spatial alignment between the 3D geometry and the images. Our approach introduces a novel image selection algorithm based on ray-tracing visibility analysis and integrates optimization techniques such as occlusion-aware texture mapping, texture atlassing, and optimization methods. Experimental results demonstrate the tool’s effectiveness in producing high-quality, visually enriched 3D models, significantly reducing the time and effort required for manual processing. Proposed optimization methods reduce the number of textures by 97%, the total texture size by 63%, and the CityJSON file size by 37% relative to the non-optimized model. Although this study focused on one urban area, the methodology is scalable to large city datasets and suitable for national-level applications involving thousands of buildings. Visualization of textured model generated using developed method is available at: https://bit.ly/textured3dbag .