Large deformation measurement using digital image correlation: a fully automated approach.

Area of Science:

• Mechanical Engineering
• Computer Vision
• Materials Science

Background:

• Digital image correlation (DIC) relies on iterative cross-correlation for accurate point matching.
• Accurate initial guesses of deformation parameters are crucial for DIC algorithm convergence.
• Large rotations and heterogeneous deformations pose challenges for traditional DIC methods.

Purpose of the Study:

• To develop a fully automated method for accurately initializing deformation parameters in DIC.
• To address challenges posed by large rotations and heterogeneous deformations in image analysis.
• To improve the convergence speed and accuracy of DIC algorithms.

Main Methods:

• Utilizing a robust computer vision technique to match feature points between reference and deformed images.
• Initializing seed point deformation parameters via an affine transform fitted to local feature points.
• Employing a modified quality-guided initial guess propagation scheme to update adjacent points.

Main Results:

• The proposed method provides a complete and accurate initial guess for all measurement points.
• Demonstrated rapid and correct convergence of the nonlinear optimization algorithm.
• Successfully handled deformed images with significant rotation and heterogeneous deformation.

Conclusions:

• The automated initialization method enhances the robustness and efficiency of digital image correlation.
• The technique is effective for analyzing complex deformation scenarios, validated by simulations and experiments.
• This approach facilitates more reliable quantitative analysis in experimental mechanics.