Development of Optimal Multiscale Patterns for Digital Image Correlation via Local Grayscale Variation

In many applications of digital image correlation (DIC), it is advantageous to have measurements at multiple scales. Because it is rare to have natural features that can be used for DIC at multiple magnifications, an appropriately multiscale DIC pattern is needed. This work develops a multiscale DIC pattern that (1) contains features appropriate for both high and low magnification, (2) does not need to know the location of high magnification a priori, and (3) does not require specialized DIC equipment beyond what is necessary to achieve the two magnifications. The pattern is developed based on an optimization framework that minimizes expected DIC error while constraining sub-regions of the pattern to biased average grayscale values. The inclusion of local grayscale biases in the pattern has the effect of introducing resolvable features at a length scale much larger than the speckles of which the pattern is composed. Numerical and physical experiments were performed to illustrate the functionality and utility of the designed patterns. Notable among the findings is the trade off between DIC accuracy at the two scales and how it is controlled by grayscale bias.