Increasing accuracy and precision of digital image correlation through pattern optimization

- The component of DIC accuracy due to an applied pattern is quantified and assessed.
- The relationship between several pattern quality metrics and accuracy is investigated.
- Optimized patterns are produced based on the pattern quality metrics.
- The optimized patterns are approximately 12% more accurate and precise than random patterns.

The accuracy and precision of digital image correlation (DIC) is based on three primary components: image acquisition, image analysis, and the subject of the image. Focus on the third component, the image subject, has been relatively limited and primarily concerned with comparing pseudo-random surface patterns. In the current work, a strategy is proposed for the creation of optimal DIC patterns. In this strategy, a pattern quality metric is developed as a combination of quality metrics from the literature rather than optimization based on any single one of them. In this way, optimization produces a pattern which balances the benefits of multiple quality metrics. Specifically, sum of square of subset intensity gradients (SSSIG) was found to be the metric most strongly correlated to DIC accuracy and thus is the main component of the newly proposed pattern quality metric. A term related to the secondary auto-correlation peak height is also part of the proposed quality metric which effectively acts as a constraint upon SSSIG ensuring that a regular (e.g., checkerboard-type) pattern is not achieved. The combined pattern quality metric is used to generate a pattern that was on average 11.6% more accurate than a randomly generated pattern in a suite of numerical experiments. Furthermore, physical experiments were performed which confirm that there is indeed improvement of a similar magnitude in DIC measurements for the optimized pattern compared to a random pattern.