Camera-Based Dynamic Vibration Measurement Using CoTracker Model

This paper investigates the CoTracker model, a sophisticated deep learning approach based on transformer architecture, for its application in dynamic vibration measurement using vision cameras. Compared to traditional accelerometers, which are costly and limited to point-based measurements, CoTracker promises a cost-effective, full-field analysis by tracking pixel movements in video sequences. We applied this model to analyze the vibrations of a cantilever beam captured through a high-speed camera, comparing its output against data from a standard reference accelerometer. Our findings suggest that CoTracker could revolutionize structural vibration monitoring due to its efficiency and broad applicability.