Digital Image Correlation Using Stochastic Parallel-Gradient-Descent Algorithm

This article proposes a digital image correlation (DIC) method based on the stochastic parallel gradient descent (SPGD) algorithm. Stochastic parallel perturbations are imposed on deformation parameters to make the correlation coefficients converge to a global extremum; thus, this allows the final measured values of the deformation parameters to be obtained and the DIC measurement to be made. Both simulated and real data processing, including rigid body and strain deformation, show that the proposed method can achieve nearly the same accuracy as the Newton-Raphson (NR) method in most cases and higher accuracy in some cases, such as the simulated experiments of rigid body translation with and without noise. It also has a good noise-robustness. Furthermore, a series of experiments have been designed to evaluate the convergence characteristics of the proposed method, and it has been proved able to process large displacement and have a stable convergence process, good robustness, and a high convergence speed when bilinear interpolation is adopted.