A High-Precision Method for Extracting Lateral Deformation in Operational Shield Tunnels Based on LiDAR Point Cloud...

Area of Science:

• Civil Engineering
• Geotechnical Engineering
• Computer Vision

Background:

• Structural health assessment of operational shield tunnels is crucial for urban rail transit safety.
• LiDAR point clouds from tunnels present challenges like occlusions, noise, and uneven density.
• Existing section-by-section ellipse fitting methods suffer from parameter instability.

Purpose of the Study:

• To develop a high-precision method for extracting lateral deformation from tunnel LiDAR point clouds.
• To address the limitations of conventional methods in handling noisy and complex tunnel environments.
• To improve the stability and accuracy of deformation monitoring in operational shield tunnels.

Main Methods:

• A point-wise attention Transformer network (PWAT) was developed for accurate lining segmentation.
• PWAT utilizes k-NN adaptive sampling, geometric position encoding, and geometry-constrained multi-head self-attention.
• A continuity-constrained RANSAC (CC-RANSAC) algorithm was introduced to enhance ellipse parameter stability.

Main Results:

• PWAT achieved 99.53% overall accuracy and 99.06% mIoU in six-class segmentation.
• CC-RANSAC reduced the mean residual to 2.0 mm and the center jump rate to 4.2%.
• The method demonstrated a mean absolute error of 1.35 mm and RMSE of 1.68 mm compared to total station data.

Conclusions:

• The proposed PWAT and CC-RANSAC method enables automatic and accurate extraction of lateral deformation from tunnel LiDAR point clouds.
• This approach significantly improves the stability and precision of deformation monitoring in operational shield tunnels.
• The findings contribute to enhanced structural health assessment for urban rail infrastructure.