Structural modal identification based on motion magnification and improved subpixel edge detection algorithm

Modal identification is essential to structural health monitoring (SHM). The use of computer vision (CV) for modal identification offers several advantages compared to conventional approaches. However, most existing CV technologies require the installation of artificial targets and are limited to scenarios with significant vibration amplitudes. Therefore, this paper proposes a target-free modal identification approach integrating the broad-band phase-based video motion magnification (BPVMM) technique with an improved subpixel edge detection technique. The BPVMM technique is employed to amplify the low-amplitude vibrations and the improved subpixel edge detection algorithm based on partial area effect (PAE) is applied to extract displacement from the amplified video. The data-driven stochastic subspace identification (Data-SSI) method is then used to identify the frequencies and mode shapes of the structure. Finally, the performance of the developed technique is evaluated through laboratory and field tests to verify its accuracy. The results show that the mode shapes identified by the proposed technique are in good agreement with the reference mode shapes, and the MAC value is above 0.9843. This study provides a convenient, no-target, and low-cost method for structural modal identification.