On the computer vision-based approach for output-only wave-based structural identification and health monitoring

A non-contact (vision-based) output-only technique is explored for monitoring and health assessment of civil structures such as multistory buildings and bridges. A consumer-grade camera is used to record the vibrations under natural ambient excitation. Then, the Kanade-Lucas-Tomasi (KLT) algorithm is utilized to track the displacements of desired target points (i.e., the points representing the location of virtual receivers located along the height of the building) and extract the pixeldomain time histories. In the next step, the extracted time history responses are deconvolved between pairs of target points to extract the Impulse Response Functions (IRFs) of each pair. This process, often known as seismic interferometry, extracts the causal and acausal wave propagation behavior between two points of the structure. Depending on the frequency content, this behavior is of interest to reveal structural damage along wave propagation paths. The technique is conceptualized by a numerical example administrated in ABAQUS. It is followed by a preliminary test conducted on a small-scale multistory frame structure subjected to a random shaker excitation, which is monitored by a cellphone’s camera. The main current challenge of this technique is the low frequency of the image recording from the camera. The next phase of this work will utilize a high-speed camera to extract IRFs at frequencies high enough for localized damage detection.