Estimation of cable tension in large-span cable-stayed bridges using image super-resolution reconstruction technology

Abstract

This study proposes a new method for remotely estimating cable tension in large-span cable-stayed bridges, addressing challenges such as image blurring, cable vibration displacement extraction, and camera vibrations. First, the Real-Enhanced Super-Resolution Generative Adversarial Network (Real-ESRGAN) is used to enhance the image quality and edge clarity of cable-stayed bridge cables. An adaptive threshold Canny-Hough operator is then introduced to detect cable edges, reducing the uncertainty introduced by manually setting thresholds in the Canny operator and effectively minimizing interference from complex backgrounds. Next, the Perspective-four-Point camera calibration algorithm is used to determine the scaling factor, with the cable's anchorage point serving as a reference for calculating its actual vibration displacement, thereby reducing the impact of camera vibrations. Finally, Fourier transform is applied to analyze the displacement data and estimate cable tension based on vibration frequency. The method is validated through laboratory tests on a simply supported beam and field tests on a large-span cable-stayed bridge. The results demonstrate high accuracy compared to accelerometer measurements.