ABSTRACT
Computer vision-based methods provide dense and accurate structural displacement measurements, countering the limitations of traditional contact-type sensors. However, storing the extensive data generated by full-field measurements, where each pixel acts as a sensor, poses a challenge. To address this, an optimal sensor selection strategy is proposed. First, the improved phase-based optical flow (IPBOF) method, which is resistant to illumination changes and phase wrapping, is used for displacement measurement. Subsequently, the obtained displacement matrix is decomposed into tailored bases through Singular Value Decomposition (SVD), and QR decomposition is then applied to determine the best pixels for reconstruction. The efficacy of the proposed method was verified using a laboratory video capturing the vibration of a 6-story building model. The results validate the method’s ability to robustly reconstruct full-field dynamic displacement with high precision and a notable compression rate. This approach offers a solution to handle extensive data, retaining accurate full-field displacement information for long-term structural monitoring purpose.
