Table 2 Summary of data processing and integration workflow.

Phase 1 processing

TLS Data Import and Pre-processing: The point cloud from the Leica HDS-6200 laser scanner was imported. XYZ directions were adjusted to align with the project’s coordinate system

Photogrammetry Data Processing and Optimization: 10 trials were conducted in Agisoft Metashape and ReCap photo, with Trial 9 identified as optimal, using only 100 carefully selected iPhone images. This choice was made not based on the quantity of images but on the high accuracy and detailed wall geometry achieved while maintaining a manageable data size. To further enhance coverage and data quality, the best outputs from additional selected trials were also merged, resulting in a combined photogrammetry point cloud of 295,280,474 points

Noise Filtering and Subsampling (Photogrammetry): A Radius Noise filter was applied to the photogrammetry point cloud, reducing it to 197,753,019 points. Subsequently, a Distance-based Subsampling Tool (with a distance of 0.001) was used to remove duplicated points, resulting in 31,439,929 points

Alignment and Merging: Both the processed TLS point cloud (18,113,861 points) and the subsampled photogrammetry point cloud were imported and aligned. The ICP algorithm was used for precise registration. The final merged Phase 1 point cloud contained 49,553,790 points

Phase 2 processing

Drone Photogrammetry Processing: Images from the DJI Mavic 2 Enterprise Dual drone were processed using Agisoft Metashape, resulting in a 100% complete photogrammetry model. The initial point cloud, including the church and its environment, contained 78,482,919 points. This was then cleaned to isolate only the church, yielding 14,975,218 points

Supplemental TLS Data Processing: The point cloud from the Leica BLK360 (G2) laser scanner was imported, and its XYZ directions were adjusted. This initial scan contained 106,986,856 points. A Radius Noise filter was applied, reducing the count to 76,298,656 points. Further subsampling (Distance-based, 0.001) was performed to remove duplicates, resulting in 74,098,714 points

Alignment and Merging (Phase 2): The cleaned drone photogrammetry point cloud and the processed supplemental TLS point cloud were imported and aligned using ICP. The final merged Phase 1 point cloud contained 89,073,932 points

Final hybrid integration

The final comprehensive HBIM model was created by integrating the merged point clouds from both Phase 1 and Phase 2. This involved:

Initial Alignment: The Phase 1 merged cloud and the Phase 2 merged cloud were aligned using manual registration followed by ICP refinement. For the Phase 1 vs Phase 2 C2C comparison, the Phase 2 merged point cloud was chosen as the reference

Statistical Filtering: Overlapping points between the two-phase datasets were filtered to reduce noise and ensure seamless transitions

Model Merging: The combined dataset formed the final hybrid model, leveraging the accuracy of ground-based data with the completeness offered by aerial and supplemental scans, particularly in occluded and elevated regions. The final merged dataset contained 92,523,738 points