Computer vision technique for blind identification of modal frequency of structures from video measurements

Operational modal analysis (OMA) is required for the maintenance of large-scale civil structures. This paper developed a novel methodology of non-contact-based blind identification of modal frequency of a vibrating structure from its video measurement. The developed methodology uses computer vision techniques and signal separation techniques for modal frequency identification. There are two stages in the proposed methodology, first stage is extracting the motion data of the vibrating structure from its video using a multiscale decomposition computer vision technique known as complex steerable pyramid. Second stage consists of a statistical technique popularly known as principal component analysis (PCA) for dimension reduction on the motion data extracted from the video and a signal separation technique based on Hilbert transform known as analytical mode decomposition (AMD) for separating the modal frequencies. The second stage of methodology is validated by a 10-DOF numerical model. The proposed methodology is applied on the real-life video of the London Millennium bridge and an accuracy of 99% is achieved in identifying the modal frequency. This paper proves that the results of benchmark experiments reveal the competence of the recommended technique in blindly extracting the modal frequencies of the structure precisely in a non-contact computer vision-based measurement.