Reconstruction and Efficient Visualization of Heterogeneous 3D City Models

Creative Commons License

Buyukdemircioglu M., KOCAMAN GÖKÇEOĞLU S.

REMOTE SENSING, vol.12, no.13, 2020 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 12 Issue: 13
  • Publication Date: 2020
  • Doi Number: 10.3390/rs12132128
  • Journal Name: REMOTE SENSING
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Agricultural & Environmental Science Database, CAB Abstracts, Compendex, INSPEC, Veterinary Science Database, Directory of Open Access Journals
  • Hacettepe University Affiliated: Yes


The increasing efforts in developing smart city concepts are often coupled with three-dimensional (3D) modeling of envisioned designs. Such conceptual designs and planning are multi-disciplinary in their nature. Realistic implementations must include existing urban structures for proper planning. The development of a participatory planning and presentation platform has several challenges from scene reconstruction to high-performance visualization, while keeping the fidelity of the designs. This study proposes a framework for the integrated representation of existing urban structures in CityGML LoD2 combined with a future city model in LoD3. The study area is located in Sahinbey Municipality, Gaziantep, Turkey. Existing city parts and the terrain were reconstructed using high-resolution aerial images, and the future city was designed in a CAD (computer-aided design) environment with a high level of detail. The models were integrated through a high-resolution digital terrain model. Various 3D modeling approaches together with model textures and semantic data were implemented and compared. A number of performance tuning methods for efficient representation and visualization were also investigated. The study shows that, although the object diversity and the level of detail in the city models increase, automatic reconstruction, dynamic updating, and high-performance web-based visualization of the models remain challenging.