Dynamic characteristics of multi-degree-of-freedom frame systems derived from vision-based displacement measurement using consumer-grade camera video

Abstract The structural health monitoring and maintenance of large-scale buildings and infrastructure have traditionally relied on attached sensor systems. However, as structures grow increasingly complex and extensive, the number of required sensors has multiplied, resulting in higher costs and challenges in maintaining sensor network reliability. This study proposes a method for extracting dynamic characteristics from multi-degree-of-freedom (MDOF) structures through non-contact, video-based displacement measurement, using consumer-grade cameras. Specifically, a four-story building model was constructed and subjected to single-axis shaking table tests, with motion data captured via smartphone and action cameras. Experimental variables included frame rates (30, 60, and 120 fps), resolution levels (HD, FHD, UHD), and five distinct excitations. Image processing was applied to the captured videos to derive the response displacements of the test model, followed by Fast Fourier Transform (FFT) analysis to determine natural frequencies, which were subsequently compared with accelerometer measurements. To enhance the accuracy of frequency identification, a frequency detection system was developed, which sets a threshold for peak points in the normalized frequency response function. By clustering these peaks by frequency, a quantitative standard was established for identifying natural frequencies. The results showed that the root mean square error (RMSE) between the response displacement and the shaking table input displacement was maintained below 0.26 mm. In addition, natural frequencies from the first to fourth modes derived using the proposed method closely matched those obtained from accelerometer data. Under short-period excitations, the video-based approach demonstrated high reliability, achieving a Modal Assurance Criterion (MAC) exceeding 0.8 for all modes, thereby validating its effectiveness for precise dynamic characterization in structural health monitoring applications.