| Title |
A Study on Accuracy Improvement in Scan-to-BIM for Interior Inverse Design Through Area-Based Editing and Matching Methods |
| Authors |
박하준(Park, Ha-Jun) ; 유무영(Yoo, Moo-Young) |
| DOI |
https://doi.org/10.5659/JAIK.2026.42.6.121 |
| Keywords |
Reverse Design; Scan to BIM; Editing; Matching; Accuracy |
| Abstract |
This study quantified the effect of zone-specific editing and registration on indoor Scan-to-BIM accuracy while keeping all other variables
strictly controlled. Data were collected using the same device, the Trion Pl, and identical preprocessing steps, including a six millimeter voxel
grid, statistical outlier removal, and normal recalculation. As a result, any observed differences in outcomes can be attributed solely to the
editing and registration procedures applied. Three registration workflows were defined and applied according to zone characteristics. Distance
Bias Correction (DBC) was used to address scale deviations along long linear paths. Manhattan Alignment (MAN) was applied to standardize
the coordinate system and improve surface planarity in wide interior spaces. In addition, Segmentation with Loop Closure (SEG/LOOP) was
used to reduce cumulative drift in both extended and vertically complex paths. These methods were assigned by area, with DBC and
SEG/LOOP used in the corridor, SEG/LOOP in the stairwell, MAN in the wide interior, and MAN combined with a short rigid-body ICP in
the general room. Accuracy was evaluated by comparing each model against ground truth derived from direct field measurements. The
length-weighted P9S error decreased by approximately 61 percent in the corridor, 55 percent in the stairwell, and between 47 percent and 53
percent in both the wide interior and the general room. In addition, surface shape indices, including M3C2 and C2C, showed reductions
ranging from 37 percent to 61 percent at P9S. Importantly, all zones satisfied the domestic location accuracy standards established by the
Ministry of Land, Infrastructure and Transport. Overall, these results indicate that selecting editing and registration strategies based on
zone-specific error mechanisms can substantially reduce high-percentile errors and improve compliance with national quality requirements. |