A monocular vision-based dynamic response measurement method for bridge structures using homography matrix transformation

Abstract

Dynamic response analysis is critical for understanding the loads experienced by bridge structures during operation, offering important insights into their safety and stability. Accurate measurement of these responses is a cornerstone of structural health monitoring. However, traditional accelerometers, which are widely used for dynamic response measurement, may suffer from performance degradation and measurement drift over time, requiring frequent calibration and limiting their applicability for long-term monitoring. Consequently, this study proposes a monocular vision-based dynamic response measurement method for bridge structures using homography matrix transformation. By exploiting the geometric properties of circular markers, the proposed method efficiently identifies and tracks feature targets with high precision. The homography matrix facilitates a robust mapping between pixel and physical coordinates, allowing accurate measurement of dynamic accelerations and frequency-domain characteristics via Fourier decomposition. The proposed method is experimentally validated against a reference accelerometer using a laboratory-scale bridge model under variations in lighting, measurement angles, and impact loads. The results demonstrate strong agreement between the proposed method and the reference accelerometer. In the time domain, the proposed method achieves a correlation coefficient ( R2) of up to 0.997, a root mean square error below 0.198, and a maximum average relative error in acceleration amplitude of 3.494%. In the frequency domain, the main frequencies are consistent, with a maximum relative amplitude error of 2.055%. These findings demonstrate the reliability and accuracy of the proposed method, indicating its potential as a robust and cost-effective alternative for dynamic response measurement of bridge structures.