Rigid-array constraints and matching-quality-weighted global bundle adjustment for large-scale 3D-DIC reconstruction.

Area of Science:

• Metrology
• Computer Vision
• Mechanical Engineering

Background:

• Multi-step 3D Digital Image Correlation (3D-DIC) struggles with error accumulation in large-scale surface reconstruction.
• Local sequential stitching methods are prone to cumulative errors, limiting accuracy for complex geometries.

Purpose of the Study:

• To develop an advanced framework for accurate 3D-DIC reconstruction of large-scale complex surfaces.
• To mitigate error accumulation inherent in traditional multi-step reconstruction techniques.

Main Methods:

• Proposed a matching-quality-weighted global bundle adjustment framework for rigid multi-camera arrays.
• Utilized the constant relative poses of cameras in an array for reduced-dimensional parameterization.
• Integrated all scanning steps and camera observations into a single global optimization problem.
• Incorporated zero-mean normalized sum of squared differences (ZNSSD) for observation weighting to down-weight low-quality matches.

Main Results:

• Reduced stitching consistency Root Mean Square (RMS) error in overlapping regions from 0.716 mm to 0.1060 mm on a large-aperture parabolic antenna.
• Achieved a mean reprojection error (MRE) of 0.045 pixels, indicating high reconstruction precision.
• Demonstrated method robustness under extreme temperature variations (-80°C to 100°C) during vacuum thermal deformation experiments.

Conclusions:

• The proposed matching-quality-weighted global bundle adjustment framework effectively addresses error accumulation in multi-step 3D-DIC.
• The method offers superior stitching consistency and accuracy for large-scale complex surface reconstruction compared to existing techniques.
• The framework exhibits robustness and reliability, even under challenging environmental conditions.