This study integrates OGC CityGML LOD-1 (Level of Detail) building models with airborne LiDAR (Light Detection and Ranging) data to refine the models to be conforming to the LOD-2 requirements. For reconstructing LOD-2 flat-roof building models, this study develops an automatic algorithm based on a majority operator to extract the heights of stories from point cloud data. Subsequently, the elevation of building models can be adjusted to be close to real height. For pitched roof buildings, a semi-Automatic procedure is developed to adjust the elevation of building models vertex by vertex. Experimental results indicate that the averages and standard deviations of elevation differences, and the Root Mean Squared errors (RMSE) of the constructed models are 0.712, 0.489 and 0.964 meters, respectively. Further exploring the reconstructed results reveals an inconsistent building height criterion between the generated models and reference data. The elevation of a generated model is determined as the height of the roof floor (because of the majority of point cloud) from ground. However, the reference data is measured from ground to the top of parapet walls for better mapping texture façades using in-situ images, representing different viewpoints between data-Acquisition process, geometric modeling and visualization. However, the quantitative validation results, from a technical perspective, are reasonable and conform to OGC CityGML LOD-2 building model requirements.
|Number of pages||6|
|State||Published - 2018|
|Event||39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018 - Kuala Lumpur, Malaysia|
Duration: 15 Oct 2018 → 19 Oct 2018
|Conference||39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018|
|Period||15/10/18 → 19/10/18|
- 3D building model
- Level of detail
- OGC CityGML
- Point cloud.