Scholarlink

<p>Accurate displacement monitoring is essential for evaluating the structural health of civil infrastructure, yet it poses significant challenges for bridges with complex, high-frequency dynamic behavior. This research conducts a rigorous frequency-domain evaluation of a non-contact monitoring system that utilizes the advanced EF-RAFT optical flow algorithm on a short-span cable-stayed bridge. By comparing its output with synchronized accelerometer data across eight case studies, the study precisely defines the system's operational limits. The vision-based method effectively captures low-frequency global displacements (0–30&#xa0;Hz), with spectral peaks consistently within 0.2&#xa0;Hz of reference sensor data. However, a fundamental constraint is its inability to resolve high-frequency components (40–400&#xa0;Hz) due to the <i>Nyquist</i> limit. The fidelity of the computer vision system is not inherent but is highly dependent on three key parameters: frame rate, camera-to-target distance, and environmental exposure. wind-induced camera shake identified as a dominant noise source, severely degrading the Signal-to-Noise Ratio and introducing spurious spectral peaks. The findings confirm that computer vision and accelerometers are complementary technologies; the former provides global displacement data, while the latter delivers localized, high-frequency vibrations. This work concludes that a hybrid monitoring approach is indispensable for a comprehensive structural assessment, establishing a clear framework for integrating computer vision by delineating its specific capabilities and limitations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Computer vision-based structural monitoring of short-span cable-stayed bridges: a case study

  • 期刊论文

  • 发表日期：

    2025年11月18日

查看全文
  • Mojtaba Razavi,
  • Saman Yaghmaei-Sabegh,
  • Ali Hadidi

摘要

Accurate displacement monitoring is essential for evaluating the structural health of civil infrastructure, yet it poses significant challenges for bridges with complex, high-frequency dynamic behavior. This research conducts a rigorous frequency-domain evaluation of a non-contact monitoring system that utilizes the advanced EF-RAFT optical flow algorithm on a short-span cable-stayed bridge. By comparing its output with synchronized accelerometer data across eight case studies, the study precisely defines the system's operational limits. The vision-based method effectively captures low-frequency global displacements (0–30 Hz), with spectral peaks consistently within 0.2 Hz of reference sensor data. However, a fundamental constraint is its inability to resolve high-frequency components (40–400 Hz) due to the Nyquist limit. The fidelity of the computer vision system is not inherent but is highly dependent on three key parameters: frame rate, camera-to-target distance, and environmental exposure. wind-induced camera shake identified as a dominant noise source, severely degrading the Signal-to-Noise Ratio and introducing spurious spectral peaks. The findings confirm that computer vision and accelerometers are complementary technologies; the former provides global displacement data, while the latter delivers localized, high-frequency vibrations. This work concludes that a hybrid monitoring approach is indispensable for a comprehensive structural assessment, establishing a clear framework for integrating computer vision by delineating its specific capabilities and limitations.