Structural Low-Level Dynamic Response Analysis Using Deviations of Idealized Edge Profiles and Video Acceleration Magnification

The development of high-speed camera systems and image processing techniques has promoted the use of vision-based methods as a practical alternative for the analysis of non-contact structural dynamic responses. In this study, a deviation extraction method is introduced to obtain deviation signals from structural idealized edge profiles. Given that the deviation temporal variations can reflect the structural vibration characteristics, a method based on singular-value decomposition (SVD) is proposed to extract valuable vibration signals from the matrix composed of deviations from all video frames. However, this method exhibits limitations when handling low-level motions that reflect high-frequency vibration components. Hence, a video acceleration magnification algorithm is employed to enhance low-level deviation variations before the extraction. The enhancement of low-level deviation variations is validated by a light-weight cantilever beam experiment and a noise barrier field test. From the extracted waveforms and their spectrums from the original and magnified videos, subtle deviations of the selected straight-line edge profiles are magnified in the reconstructed videos, and low-level high-frequency vibration signals are successfully enhanced in the final extraction results. Vibration characteristics of the test beam and the noise barrier are then analyzed using signals obtained by the proposed method.