A Geometric-Feature-Based Method for Automatic Extraction of Anchor Rod Points from Dense Point Cloud.

Area of Science:

• Geomatics Engineering
• Civil Engineering
• Geotechnical Engineering

Background:

• High-precision 3D laser scanning is vital for hydraulic building modeling and deformation monitoring.
• Point cloud quality is critical for accurate data processing in these applications.
• Anchor rod points and mixed pixels can impact modeling and monitoring accuracy.

Purpose of the Study:

• To develop an automatic method for extracting anchor rod points from high-density point clouds.
• To improve the quality and reliability of point cloud data for construction caverns.
• To address the challenges posed by anchor rod points and mixed pixels in underground powerhouses.

Main Methods:

• A two-step workflow for automatic anchor rod point extraction from high-density point clouds.
• Utilizing triangular mesh features and anchor rod parameters for initial identification in downsampled data.
• Employing curvature features for precise anchor rod point extraction.

Main Results:

• Initial identification accuracy of 97.2% for anchor rod points.
• Precise extraction method based on curvature curve fitting is effective.
• Accurate separation of three anchor rod types from dense point clouds on rough cavern roof surfaces.
• Low false-extraction rates ranging from 0.11% to 5.09%.

Conclusions:

• The developed method provides high-quality and dependable data for point cloud registration, modeling, and deformation analysis in construction caverns.
• This approach enhances the reliability of data used in hydraulic engineering projects.
• The automatic extraction of anchor rod points significantly improves data processing efficiency and accuracy.