Extraction and location of subway shield tunnel segment joints from RMLS point clouds
Precisely extracting segment joints in subway shield tunnels is critical for automated safety evaluation such as segment assembly quality inspection, dislocation and deformation detection and lining surface defects diagnosis. However, traditional extraction methods can only extract segment joints of...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-08-01
|
| Series: | International Journal of Digital Earth |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/17538947.2025.2528627 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Precisely extracting segment joints in subway shield tunnels is critical for automated safety evaluation such as segment assembly quality inspection, dislocation and deformation detection and lining surface defects diagnosis. However, traditional extraction methods can only extract segment joints of specific types (circumferential joints or radial joints) or specific assembly patterns (straight joints or staggered joints) and their extraction accuracy is easily affected by nearby tunnel facilities. To meet distinct engineering requirements, a Rail-borne Mobile Laser Scanning (RMLS) point cloud-based method capable of simultaneously identifying all joints types and patterns is presented. The proposed method first employs an elliptical fitting residual statistics algorithm to remove non-lining points, eliminating their adverse effects on joint extraction. Then, cross-sections corresponding to circumferential joints are automatically identified and spatially localized using an adaptive multi-threshold algorithm based on local intensity statistics, dividing the tunnel lining into individual shield rings. Finally, a novel curvature-based ratio metric, derived from the local bulge and liner distribution characteristics, is developed to identify and localize radial joints within each shield ring. Experiment results show that the proposed method achieves average IoU, recall, and accuracy of 92.8%, 95.3%, and 94.3%, respectively, even surpassing the performance of deep learning-based semantic segmentation network. |
|---|---|
| ISSN: | 1753-8947 1753-8955 |