A Ring-Projection-Based Two-Scale Approach for Accurate Digital Image Correlation of Large Translations and Rotations

Abstract

Background

Digital image correlation (DIC) has been widely used for motion tracking and estimation, however, the process is often sensitive to the initial guess, especially under large translations and rotations.

Objective

To provide novel and effective solutions for the DIC in measuring large translations and rotations.

Methods

A ring-projection-based two-scale approach is proposed. In the integer-pixel scale, a novel ring projection scheme, including amplitude and phase correlations of the rings, is developed to quickly get the integer-pixel initial estimation of the translations and rotation. In the sub-pixel scale, the gradient-based inverse compositional Gauss-Newton (IC-GN) algorithm, which is free from repeat computation of Hessian matrix, is adopted to efficiently get the optimal motion parameters.

Results

The numerical example show that the absolute error is no more than 0.05 pixel for measured large translations and no more than 0.05\(^\circ\) for measured large rotations. While test experiments on a rotated blade and a flexible arch demonstrate the effectiveness, accuracy and applicability of the proposed approach in measuring the rotating motion, flexible large deformation and vibrational modal parameters of structures.

Conclusions

The ability and effectiveness of the proposed approach for large translations and rotations measurement have been verified. Since large deformations and rotations are frequently encountered in rotating and flexible structures, the proposed approach is believed to constitute a feasible and powerful tool for static and dynamic deformation measurement of these structures.