Automated defect detection in FRP-bonded structures by Eulerian video magnification and adaptive background mixture model

Abstract

A computer-aided methodology for automated defect detection in fiber-reinforced polymer (FRP) bonded civil engineering structures via Eulerian video magnification integrated with adaptive background mixture model is demonstrated. In this methodology, Eulerian video magnification is firstly used to render the stimulated de-bonding motion visible in a high-speed and high-resolution video. Then, adaptive background mixture model is applied in the motion-magnified video for automated tracking of the de-bonding motion. The combined use of these two computer-aided vision techniques aims at developing an innovative way for intuitive, straightforward, and automated defect detection. In the application of this methodology, two operational parameters (i.e. motion amplification factor and de-noising factor) in the video processing can greatly affect the sensitivity of defect detection. After evaluation of their effects, the present work shows a practice guide for adjusting the above two influencing parameters in order to improve the automated defect detection.