.png){kind=icon}
.png){kind=icon}
Return
Semi-automatic TLS–HLS point cloud registration using reflection-induced void features for structural diagnostics
.png){kind=icon}
OA
AI
DOI：10.1016/j.measurement.2026.122930.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
•
A semi-automatic framework for TLS–HLS point cloud registration is presented.
•
The ATHIR algorithm enables automatic detection of target locations using TLS point-cloud voids.
•
The approach reduces manual intervention in target-based registration workflows.
•
Sub-millimetre integration accuracy supports reliable mapping of local structural defects.
•
The method ensures metrological consistency in multi-scale point cloud integration.
Abstract
Terrestrial Laser Scanning (TLS) is widely used for large-scale geometric documentation of structures, while handheld laser scanning (HLS) provides high-resolution data for detailed inspection of local features. The registration of these datasets is essential for multi-scale structural diagnostics; however, it remains challenging due to differences in spatial scale, measurement precision, and coordinate systems.
This paper presents a method for the semi-automatic registration of HLS data within a TLS reference frame using reflection-induced void features. The approach exploits the absence of points in TLS data caused by specular reflection from circular targets, enabling their identification and precise centroid estimation. The detected target centroids are subsequently matched with corresponding HLS reference points, and a rigid transformation is computed to preserve the metrological integrity of the high-resolution data. The proposed workflow combines geometric normalization, voxel-based segmentation, circle fitting, and robust correspondence estimation using distance invariants and RANSAC filtering. The method operates automatically on preselected regions of interest and does not require manual identification of corresponding points.
Experimental validation performed on a real building structure demonstrates that the proposed approach enables reliable registration of multi-scale point clouds, achieving a root mean square error (RMSE) of approximately 0.6–0.7 mm. The results confirm that high-resolution HLS data can be accurately integrated into a global TLS reference frame while maintaining sub-millimetre geometric fidelity. The presented method provides a metrologically consistent solution for target-based TLS–HLS data registration and can support detailed structural diagnostics in engineering applications.
Keywords:
Semi-automated point cloud registration
TLS
HLS
Rigid transformation
Multi-scale measurements
Structural inspection
ATHIR
,
Automated TLS-HLS integrated registration
BIM
,
Building information modelling
CCL
,
Connected component labelling
ETC
,
Extracted target centroids
HLS
,
Handheld 3D laser scanning
ICP
,
Iterative closest point
k-d tree
,
k-dimensional tree
PCA
,
Principal component analysis
RANSAC
,
Random sample consensus
RMSE
,
Root mean square error
SVD
,
Singular value decomposition
TLS
,
Terrestrial laser scanning
VGP
,
Voxel-grid partitioning
SHM
,
Structural health monitoring
ROI
,
Region of interest
Journal
IF:
5.6
Papers:
2.0W
Citations:
5.4W
View more organizations