Surface crack detection in concrete structures using video processing techniques

Surface crack patterns are one of the earliest damage signs in concrete structures. Existing procedures to visually evaluate the damage rely on experts' judgment to interpret the existing cracks. The initial necessary step to quantify and automate this procedure is crack detection. Precise crack detection provides a reliable basis to update the structural parameters and to predict future behavior. Several methods have been investigated to detect cracks based on image processing methods; but, there are several limitations and inaccuracies in these methods. In a number of cases, recordings during damage occurrence are available. The videos comprise not only spatial information but also temporal information. The videos provide a set of images for a unique damage situation. In this study, using video processing methods, a methodology is developed to track crack formation. In this regard, robust principal component analysis is employed to detect new crack propagation. The experimental test data of RC shear walls are used to assess the implemented methodology. The quasi-static cyclic load is applied to these walls, and several cameras captured the video of walls' behavior. Taking advantage of the phase-based motion processing method, a video stabilization is implemented to enhance the accuracy of the crack detection step. Propagation of cracks is monitored by calculating Gini coefficients for each frame. The results show that monitoring this coefficient can indicate new crack formations.