Preprocessing methods for reducing the impact of image decorrelation in laser speckle digital image correlation

The digital image correlation method based on laser speckle pattern (LS-DIC) can achieve non-contact in-situ strain measurement of materials under complex and extreme conditions. However, laser speckle patterns are significantly affected by environments, which lead to image decor relation and reduce the accuracy of DIC calculations. To address this issue, we propose a preprocessing method for image sequences that employs the speed-up robust features (SURF) algorithm to match selected corresponding subsets within images and calculate the zero normalized cross-correlation (ZNCC) values of these subsets. By comparing the ZNCC values with a predefined threshold, it is possible to filter out decor related images from the sequence and simultaneously update reference images in real time. The efficacy of this preprocessing technique was corroborated through thermal strain detection experiments on aluminum specimens, demonstrating its capability to mitigate image decor relation effects caused by heat haze, abrupt environmental changes, and vibrations. The Pearson correlation coefficient (PCC) was utilized to quantitatively assess the variation in DIC measurement precision before and after image preprocessing. Results indicate that filtering and removing decor related images from the sequence can improve the DIC computational precision by 11%, thereby confirming that this preprocessing method can effectively improve the robustness and accuracy of LS-DIC measurements under complex and extreme conditions. (c) 2025 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement