Multi-dimensional and multi-temporal motion estimation of a beam surface during dynamic testing using low-frame rate digital cameras

Abstract

The combination of digital close-range photogrammetric systems and image processing techniques has been employed in structural health monitoring applications for more than 10 years. The use of off-the-shelf entry level digital single-lens reflex cameras has lately become a suitable choice even for applications requiring sub-millimetre- level precision especially when the involved devices need to be inexpensive. The drawback of such low-cost cameras is in the trade-off between spatial resolution, frame rate, and burst rate—at the highest available spatial resolution, a high frame rate is either not possible or it has a low burst rate. This may be problematic when monitoring a structural component during a dynamic/fatigue test. In order to estimate specimen motion in such a situation, this paper proposes an innovative sinusoidal fitting based on a least squares adjustment. This method simultaneously processes multiple bursts of data in order to synthetically increase the sampling frequency of the system. The input data for the adjustment comes from a full surface modelling procedure based on a newly proposed generalized 3D polynomial. The experimental results include a beam deformation test performed in a structures laboratory. The new sinusoidal fitting method effectively increased the system temporal resolution three-fold, which improved the precision of the derived parameters with up to two orders of magnitude. The root mean square error of the residuals were as good as 26 μm, and the one of the estimated amplitudes from the photogrammetric system versus a set of laser transducers used as control was as small as 43 μm.