Computer vision-based dynamic identification of a reinforced concrete elevated water tank

There is a growing need for monitoring the structural health conditions of aging structures and for prioritizing maintenance works to extend their safe service life. This requires cheap, flexible, and reliable tools suitable for everyday use in engineering practice. This paper presents a computer vision-based technique combining motion magnification and statistical algorithms to calculate structural natural frequencies under environmental noise excitation, and its application to a reinforced concrete elevated water tank. Digital videos were recorded from various standpoints and post-processed by tracking in time either the variation of the grey-intensity or the motion of selected pixels. Computer vision-based outcomes were validated against accelerometric measurements and integrated to them to improve the understanding of the dynamic behaviour of the water tower, which, counterintuitively, resulted anything but trivial to predict.