Marker-free full-field approach for structural 3D vibration response measurement based on Gaussian process regression and pyramidal optical flow

Abstract

Vision-based measurement techniques are drawing attention due to their non-contact, flexible and efficient advantages. However, most existing techniques require artificial markers for stereo matching and monocular tracking. The marker-free techniques are difficult to achieve full-field stereo matching and currently only applicable to 2D vibration measurements. This paper proposes a novel marker-free full-field approach for 3D vibration response measurement. Firstly, stereo matching is considered a regression problem, and Gaussian process regression (GPR) is used to model the relationship between matching point pairs. With the dataset provided by feature point matching, marker-free stereo matching of full-field arbitrary points is efficiently achieved using trained GPR. Secondly, the pyramid optical flow is applied to marker-free visual tracking, which can simultaneously track the full-field matching point pairs in two video sequences without artificial markers. Finally, the 3D vibration responses are obtained by stereo camera calibration and 3D reconstruction. The accuracy and performance of the proposed approach are evaluated through vibration measurement tests on a thin-walled pipe. The vibration responses measured by the proposed approach are compared with those measured by the laser vibrometer, accelerometer, and digital image correlation approach. Besides, a modal analysis test on a thin plate is constructed in which the full-field response measured by the proposed approach is used as the input of the modal analysis. The results demonstrate that the proposed approach can accurately measure the 3D vibration responses. And it has excellent capability in marker-free full-field vibration measurements and great potential in modal analysis.