Recent years have seen significant advancements in the field of structural health monitoring (SHM) using computer vision, which has steadily developed into a practical and effective technique for measuring the dynamic properties of structures. With the long-range, non-contact, and easy-to-use features of this measurement technology, some difficulties associated with utilizing conventional techniques to detect the natural frequencies of tall structures can be reduced. Because it is label-free, the method also avoids attaching physical targets or sensors to energized, difficult-to-access structures. This paper proposes a label-free computer vision measurement method, which uses a label-free detection system to determine the tracking feature points, and then tracks the feature points of the structure based on the KLT optical flow method. Finally, the natural frequencies of the structure are obtained by frequency domain analysis of the vibration signal of the feature points. Experimental studies on indoor transmission tower models and outdoor high street lamps were conducted to confirm the viability of the above technology. The test findings were compared with the measurements from a 941B accelerometer, and the impact of several aspects, such as frame rate, resolution, and measuring distance, on the accuracy of the results was examined. The test findings show that the label-free system identifies the first several natural frequencies, including the fundamental, with the modal-frequency estimates agreeing with the 941B accelerometer to within roughly 0.02–0.04 Hz across the first three modes; the larger percentage error at the fundamental reflects its low frequency rather than reduced accuracy, since the absolute discrepancies are comparable across all modes. The study’s findings may serve as a guide for developing software that will be used in the future to assess structural health monitoring using computer vision.
Chen et al. (Tue,) studied this question.