Panoramic deformation measurement and crack identification in concrete with deep-learning-based multi-camera DIC

Digital image correlation (DIC) has been widely used as a powerful and practical deformation measurement technique for concrete. However, due to the heterogeneity and brittleness of concrete, conventional 3D-DIC encounters significant challenges when applied to concrete materials and structures, such as limited measurement regions, difficulties in correlation analysis near cracks, and the lack of effective crack identification methods. To tackle these issues, this work proposes a novel deep-learning-based multi-camera DIC, which utilizes a multi-camera system with a face-to-face-camera-pairs configuration for panoramic/dual-surface image capture. To address the calculation challenge caused by cracks and realize pixel-wise dense yet accurate deformation field measurements for concrete, the newly proposed deep-learning-based 3D-DIC algorithm is utilized. The measurement results from discrete systems are unified into the same coordinate system using a stereo calibration block. Based on the measured panoramic pixel-wise displacement fields, the gray level residual (GLR) fields are employed for panoramic crack identification. The feasibility and accuracy of the proposed method were validated through two compression experiments of concrete samples with different shapes.