Structural Load Estimation Using Machine Vision and Surface Crack Patterns for Shear-Critical RC Beams and Slabs

This paper presents the results of using machine vision to relate surface observations to quantitative load level estimates in structural components. In particular, image-processing and machine-learning regression techniques were used to build estimation models capable of quantifying internal load levels (i.e., shear and moment) in shear-critical RC beams and slabs based on surface crack pattern images. The estimation models were generated and tested using image data sets obtained from 10 different earlier published studies, which provided 558 crack pattern images captured from 84 shear-critical RC beams and slabs across a range of load and damage levels. Working with these existing image data sets, various textural and geometric attributes of surface crack patterns were defined and evaluated with respect to their effectiveness in building useful estimation models. Multiple statistical error measures and cross-validation methods are used to quantify predictive accuracy, and several training/test methodologies were considered relative to potential field application scenarios. The results show that the estimation models based on surface crack image data can work well across a wide range of geometries, loadings, concrete strengths, and reinforcement details. Size effects can be accounted for by including specimen physical dimensions in the feature sets used for model training, and fundamental design relations can be used to develop useful nondimensional prediction parameters.