Enhancing stereo-DIC in a harsh environment via speckle-aided stereo self-calibration

Stereo digital image correlation (stereo-DIC) deployed in harsh or long-running tests often suffers from time-varying drift of camera intrinsics/extrinsics and lens distortion induced by thermal expansion, micro-motions, and vibration, which introduces systematic triangulation bias and amplifies errors in 3D displacement and strain. To address this limitation without interrupting experiments for repeated target-based recalibration, we propose a speckle-aided stereo self-calibration framework that leverages the specimen’s random speckle texture as an in-situ calibration source. Sub-pixel stereo correspondences are extracted directly from measurement images via DIC-based dense matching and are coupled with a stereo imaging model to jointly refine stereo intrinsics, extrinsics, and distortion through a robust bundle-adjustment formulation with outlier suppression and reliability screening, enabling periodic or on-demand compensation of parameter drift. The method is validated using calibration accuracy, robustness to small pose perturbations, and long-duration stability. The experimental results demonstrate that the proposed speckle-aided self-calibration provides an effective, target-free mechanism for sustaining stereo-DIC accuracy and repeatability in harsh-environment measurements.