Harnessing Digital Image Correlation and Graph Convolutional Networks for Real-Time Monitoring of Rock Deformation

Understanding rock primary deformation zones and subsequent deformation trends is crucial for guiding engineering applications. While Digital Image Correlation (DIC) can provide valuable information about the current deformation state of rocks, predicting future deformation trends is equally important to ensure the stability and efficiency of construction processes. To address this challenge, this study proposes a novel method that combines DIC with Spatio-Temporal Graph Convolutional Networks (STGCN) to predict rock deformation. Initially, feature points in the regions of interest (ROIs) of existing deformation images are identified, and these feature points between pre- and post-deformation images are matched to obtain deformation information for each point. Subsequently, graph structures are constructed, and the STGCN is applied to predict the deformation information of the feature points at subsequent time steps. Finally, the Inverse Distance Weighting (IDW) interpolation algorithm is used to obtain full-field displacement and strain, enabling real-time monitoring of rock deformation. The effectiveness of the proposed method is validated through comparative and analytical experiments on deformation prediction using oil sand images with water injection. The proposed method can not only evaluate the primary locations of deformation but also predict deformation trends, providing full-field deformation to meet the requirements of subsequent experimental analysis.