Integrating existing high resolution geospatial datasets for LoD2 building modelling: Accuracy evaluation and comparative analysis in Louvain-la-Neuve
Files
Chen_Kailu_72002400_2025.pdf
Open access - Adobe PDF
- 6.28 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Accurate representation of the built environment is essential for urban planning, analysis, and decision-making. This thesis focuses on constructing a detailed 3D building model of Louvain-la-Neuve by integrating multiple high-resolution spatial datasets. Specifically, the study assesses the compatibility between LiDAR and PICC building footprints, and examines how each dataset contributes to the generation of Level of Detail 2 (LoD2) building models. To achieve this, building footprints from both datasets were compared in terms of geometric alignment, shape similarity, and attribute completeness. The study also evaluated how differences in data resolution and classification quality influence the modeling process. A semi-automated 3D building modeling workflow was implemented using rule-based methods, and the resulting model was assessed for geometric accuracy against a LiDAR-derived digital surface model (DSM). The findings reveal notable differences in footprint precision and roof segmentation accuracy between the datasets. While LiDAR provides more detailed elevation data, PICC offers consistent and cartographically clean building outlines. Their integration improves the overall completeness of the model but introduces challenges in data fusion and height assignment. This research highlights the importance of understanding dataset interoperability when generating LoD2 building models and offers practical insights into the contributions and limitations of different data sources. The results aim to support future applications in urban analysis and sustainable city planning for Louvain-la-Neuve.