STRUCTURAL 3D MONITORING USING A NEW SINUSOIDAL FITTING ADJUSTMENT
Keywords: close range photogrammetry, dynamic loading, precise 3D reconstruction, image and point cloud processing, model-based image fitting, vertical deflections, planimetric displacements
Abstract. Digital photogrammetric systems combined with image processing techniques have been used for structural monitoring purposes for more than a decade. For applications requiring sub-millimetre level precision, the use of off-the-shelf DSLR cameras is a suitable choice, especially when the low cost of the involved sensors is a priority. The disadvantage in the use of entry level DSLRs is that there is a trade-off between frame rate and burst rate – a high frame rate is either not available or it cannot be sustained long enough. This problem must be overcome when monitoring a structural element undergoing a dynamic test, where a range of loads are cycled through multiple times a second. In order to estimate deflections during such a scenario, this paper proposes a new least-squares adjustment for sinusoidal fitting. The new technique is capable of processing multiple back-to-back bursts of data within the same adjustment, which synthetically increases the de-facto temporal resolution of the system. The paper describes a beam deformation test done in a structures laboratory. The experimental results were assessed in terms of both their precision and accuracy. The new method increased the effective sampling frequency three-fold, which improved the standard deviations of the estimated parameters with up to two orders of magnitude. A residual RMSE as low as 30 μm was attained, and likewise the RMSE of the computed amplitudes between the photogrammetric system and the control laser transducers was as small as 34 μm.