• Event-based photogrammetry is introduced for the reconstruction of flexural vibration of structures. • Framework integrating event stream processing, camera calibration, marker tracking, and 3D triangulation. • Validation against LDV measurements shows sub-millimetre accuracy with residual RMS error below 0.2 mm. • High temporal resolution without redundant data for high-frequency vibration analysis. In this paper, a novel methodology for vibration measurements of flexible structures with neuromorphic vision sensors (i.e. event cameras) is proposed and validated. Unlike conventional frame-based sensors, event cameras asynchronously record changes in scene brightness, enabling high temporal resolution with reduced data redundancy. In this study, a framework is developed for marker tracking and 3D triangulation from event recordings. The experimental setup involves four synchronized event cameras observing a cantilever beam coated with a grid of markers. The beam is excited harmonically at the first two flexural resonance frequencies, and the resulting 3D trajectories of the markers are reconstructed with subpixel accuracy. The resulting flexural deflection shapes are reconstructed through bundle adjustment of the camera recordings. The results have been benchmarked against laser vibrometer measurements, which confirmed the accuracy and applicability of the proposed measurement approach. Overall, the study demonstrates the potential of event cameras for dynamic testing in scenarios where high-speed, low-latency, low-bandwidth acquisitions are required without recording redundant data.
Baldini et al. (Wed,) studied this question.