An automatic seismic response monitoring method for building structures and equipment based on a single indoor surveillance camera is proposed in this study. The computational model for monitoring the seismic responses of structures and equipment using rigidly connected cameras is established. Combining computer vision technology and image processing methods, a seismic response monitoring process for structures and equipment is developed. An automatic seismic response monitoring is carried out by incorporating semantic segmentation in the process. The method was validated through Blender-based simulation experiments and shaking table tests, confirming its feasibility and effectiveness. The seismic response monitoring results based on shaking table tests indicate that the proposed automatic seismic response monitoring method achieves the coefficient of determination (R2) values of over 90% for the displacement responses of the structure and equipment. The relative errors in monitoring the peak displacement responses do not exceed 12%. The automatic monitoring method achieves sub-pixel accuracy and exhibits good accuracy in the frequency domain, accurately identifying the fundamental frequencies of the structure. The proposed method enables the automatic and rapid monitoring of large numbers of indoor surveillance videos after an earthquake, thereby improving post-earthquake emergency response and supporting recovery decision-making.

中文翻译：

---

基于室内监控摄像机的建筑结构及设备地震响应自动监测方法


该文提出了一种基于单个室内监控摄像机的建筑结构及设备地震响应自动监测方法。建立了使用刚性连接相机监测结构和设备地震响应的计算模型。结合计算机视觉技术和图像处理方法，开发了一种用于结构和设备的地震响应监测过程。通过在过程中加入语义分割来执行自动地震响应监测。该方法通过基于 Blender 的仿真实验和摇床测试进行了验证，证实了其可行性和有效性。基于振动台试验的地震响应监测结果表明，所提出的自动地震响应监测方法对结构和设备的位移响应达到了90%以上的决定系数（R2）值。监测峰值位移响应的相对误差不超过 12%。自动监测方法实现了亚像素级精度，在频域表现出良好的精度，准确识别了结构的基频。所提出的方法能够在地震后自动快速监控大量室内监控视频，从而提高震后应急响应并支持恢复决策。