Cables are essential components of cable-stayed bridges, and cable force is an important indicator of bridge safety and integrity. While Computer Vision (CV) technology offers a promising non-contact solution, existing methods face significant difficulties in identifying extremely small, subpixel-level cable vibrations without artificial targets, especially when video quality is degraded by severe environmental disturbances such as noise and blur. This paper aims to address this specific challenge to enable robust target-free cable force identification. Specifically, a comprehensive framework is proposed to accurately estimate subpixel cable vibrations under challenging field conditions. First, the Phase-Based Motion Magnification (PMM) technique is used to magnify small cable vibrations. Then, a Sobel Gradient Descent-Based Template Matching (SGTM) technique is specifically developed to estimate cable vibrations and frequencies from the amplified video, while also mitigating the effects of environmental disturbances and calculation errors. Finally, the boundary-constrained modal eigenvalue inversion method is used to identify cable forces. The superiority of the proposed method was validated in simulation and practical case studies of an extra-large cable-stayed bridge, named Ulan Mulun River No. 4 Bridge, located in the Inner Mongolia Autonomous Region, China. It was shown that the proposed method not only achieves high precision with 0.01 subpixel accuracy under normal conditions, but also maintains 0.01-0.30 subpixel accuracy under severe disturbances, when estimating cable vibrations. In addition, the cable frequency error was less than 0.1 Hz, and the identified cable forces deviated by below 4% using only the first-mode frequency, indicating the high engineering applicability of the proposed method.
Guo et al. (2026) studied this question.