Comparative analysis of optical flow algorithms for computer vision-based structural health monitoring using vibration data

Advancements in technology have led to the increasing use of computer vision-based modal analysis methods in the field of structural health monitoring. This development has made it necessary to design these methods as reliable and user-friendly applications. On the other hand, the limited number of studies on the comparative performance analysis of such methods stands out as a significant shortcoming in the literature. Within the scope of this thesis, four different optical flow algorithms with proven success (Lucas–Kanade, Farnebäck, Horn–Schunck, and Phase-based method) are selected and a comparative analysis of their computational times and vibration frequency estimation accuracies is conducted. For the purpose of the study, a user-friendly software interface is developed, and the aforementioned methods are experimentally tested on both a steel frame, a composite beam and a reinforced concrete frame. The structural vibration data recorded by the high-resolution camera are imported into the developed interface, converted into displacement time-series signals, and then analyzed in the frequency domain using the Fast Fourier Transform. The performance of the methods is comprehensively evaluated based on the criteria of fundamental vibration frequency accuracy and processing times.