Fast initial guess estimation for digital image correlation

TL;DR

This paper introduces the adaptive incremental dissimilarity approximations algorithm (A-IDA), a fast, reliable, and path-independent method for initial guess estimation in digital image correlation, enhancing computational efficiency especially with large subsets.

Contribution

The paper proposes A-IDA, a novel initial guess estimation algorithm for DIC that improves speed, reliability, and independence from search range and subset size compared to existing methods.

Findings

A-IDA significantly reduces computational time compared to IC-GN.

A-IDA's efficiency is less affected by search range and subset size.

The algorithm is easy to implement with adjustable thresholds.

Abstract

Digital image correlation (DIC) is a widely used optical metrology for quantitative deformation measurement due to its non-contact, low-cost, highly precise feature. DIC relies on nonlinear optimization algorithm. Thus it is quite important to efficiently obtain a reliable initial guess. The most widely used method for obtaining initial guess is reliability-guided digital image correlation (RG-DIC) method, which is reliable but path-dependent. This path-dependent method limits the further improvement of computation speed of DIC using parallel computing technology, and error of calculation may be spread out along the calculation path. Therefore, a reliable and path-independent algorithm which is able to provide reliable initial guess is desirable to reach full potential of the ability of parallel computing. In this paper, an algorithm used for initial guess estimation is proposed. Numerical and real experiments show that the proposed algorithm, adaptive incremental dissimilarity approximations algorithm (A-IDA), has the following characteristics: 1) Compared with inverse compositional Gauss-Newton (IC-GN) sub-pixel registration algorithm, the computational time required by A-IDA algorithm is negligible, especially when subset size is relatively large; 2) the efficiency of A-IDA algorithm is less influenced by search range; 3) the efficiency is less influenced by subset size; 4) it is easy to select the threshold for the proposed algorithm.