Computer vision-based quantification of updated stiffness for damaged RC columns after earthquake

Concrete surface cracks are one of the primary indicators of structural deterioration; thus, crack analysis is crucial to maintain the intact serviceability of the structural components. This research investigates the relation between earthquake-induced surface damages and stiffness degradation of RC columns. The research database comprises 422 images of 109 reinforced concrete columns tested under quasi-static cyclic loading. Using image processing methods, two visual damage indices, including the cumulative crack length and the total area of the crushed zone, are extracted from images of reinforced concrete columns. In addition to the visual indices of surface damage, supplementary information from the design and mechanical properties of the RC columns is considered available. Subsequently, three nonlinear predictive equations for the stiffness degradation estimation are derived using a combination of the collected input variables. The correlation coefficient of the predictive equation, including only visual features of damaged columns, reaches 85 percent for the test dataset and has a root mean squared error of 0.13. Although adding much information enhances the accuracy by up to 87 percent, the results demonstrate that the visual characteristics of surface damage are a potent tool for the post-earthquake damage assessment of RC columns. Thus, it can be concluded that the extracted visual damage indices can be used as an Engineering Demand Parameter for post-earthquake damage assessment of RC buildings.