Target-Free, Vision-Based System Identification of Civil Structures Using Unmanned Aerial Vehicles

Vibration-based structural health monitoring (SHM) frameworks rely upon accurately identified natural frequencies and mode shapes of structures in the field, which is critical information for damage diagnostics and model updating. Vibration-based techniques have traditionally relied on discrete contact-based sensors. Despite the success of traditional sensing modalities, challenges and limitations remain: 1) the sensors need to be placed at discrete locations, and 2) the structure needs to be accessed for instrumentation. Advancements in the fields of computer vision and robotics have facilitated the use of remote sensing technology, such as cameras and Unmanned Aerial Vehicles (UAVs), in vibration-based SHM applications to address the limitations of traditional sensors. Although UAVs have been used to monitor the dynamic response of structures, these applications have primarily relied on targets or GPS to track the structure's motion, which is not always feasible due to the large scale of civil structures. To this end, the main objective of this study is to develop an end-to-end target-free, vision-based framework for system identification of civil structures using UAVs. The proposed framework incorporates a phase-based motion estimation approach to extract the structural vibration information from videos without placing targets in the scene. For videos collected by UAVs, a correction needs to be applied to account for the camera's rigid body motion. To compensate for this motion, the framework extracts the power spectral density (PSD) plot of a static object in the scene and subtracts it from the PSDs of the structure-of-interest. To evaluate the efficacy and robustness of the developed framework, an experimental study was conducted to monitor the free vibration response of two single-degree-of-freedom structures using three different UAVs in a controlled laboratory environment. The analysis shows strong agreement between the results extracted from UAVs equipped with high-resolution cameras with those from a stationary camera and those from accelerometers. Furthermore, the results show that camera resolution, alignment, and motion can significantly impact the accuracy of the results. This study shows the potential of successfully incorporating UAVs into target-free vision-based dynamic monitoring frameworks.