Keypoint Detection-Based Real-Time Detection Method for Seismic-Induced Shaking Angles in Suspended Non-Structural Components

The functional failure of non-structural components during earthquakes severely compromises building seismic resilience. However, existing research predominantly focuses on static damage assessment of structural components, lacking effective methods for quantitative detection of dynamic seismic responses in non-structural components. To address this gap, this study proposes a keypoint detection-based real-time detection method for seismic-induced shaking angles in suspended NSCs. By integrating YOLOv8 keypoint detection with OpenCV image processing techniques, an image-based detection system was developed to dynamically quantify component shaking angles. Validation was conducted using a dual-channel dataset (”laboratory-real scenario”) constructed from shaking table tests and real-world seismic monitoring data. Experimental results demonstrate that the proposed method achieves a keypoint detection accuracy of 95.4 %, enabling precise localization of critical points in suspended components and geometric quantification of cable inclination features. This work bridges the technical gap in real-time dynamic monitoring of non-structural components and advances the transition from static analysis to real-time damage assessment, providing a robust framework for enhancing seismic resilience evaluation in modern structures.