Phase‐Based Real‐Time Vibration Monitoring and Cable Force Estimation by Computer Vision

This is the authors' abstract. We don't add key points for this paper.

Cables are critical load‐bearing components of cable‐stayed bridges, and their force condition is essential for bridge safety and serviceability. Traditional contact‐based cable force monitoring methods rely on dense sensor deployment, which faces challenges in reliability, durability, and maintenance. Therefore, computer vision (CV)‐based noncontact cable force monitoring has become an important research direction. Although existing vision‐based methods can achieve sufficient accuracy for microvibration measurement, their real‐time applicability remains limited. This paper proposes a full‐process phase‐based real‐time algorithm for cable vibration estimation and cable force identification. The proposed method uses a phase‐based direct method (PDM) to optimize adaptive high‐confidence regions and achieve subpixel motion tracking. A loopback constraint strategy is further introduced to effectively suppress cumulative drift errors during long‐term tracking. Tests on simulated and handheld‐captured videos demonstrate the reliability of the proposed framework. The method achieves accurate frequency estimation at extended distances, with minimum detectable amplitudes of 0.1–0.2 mm at 25 m, 0.3–0.8 mm at 50 m, and 1.0–1.8 mm at 100 m. In practical applications, the cable force identification errors are below 5% for most cables. Moreover, the single‐frame processing time remains below 28 ms with a moderate window size, satisfying the real‐time requirement of 30 FPS.