Decentralized operational modal identification using drone-mounted vision system via homography transformation

Due to recent advancements in camera and mobile sensing platforms, drone-based structural inspection techniques have gained considerable interest in structural health monitoring. In particular, drone-based techniques provide promising alternatives for contactless and remote inspections for hard-to-reach areas at low cost. Although these drone-assisted image-based inspections have shown significant success for fast, safe, and cost-effective structural assessments, their application in vibration-based structural monitoring is still in the infancy stage. The majority of drone-assisted vibration monitoring remains limited to lab-scale experiments due to the challenges of accurately estimating drone motion in field conditions, with many limited to the extraction of natural frequencies. The primary challenge in estimating mode shapes is capturing phase information from low-amplitude vibration under noisy field conditions. This paper presents a contactless, drone-based vision methodology for decentralized modal identification, where modal parameters are extracted using a drone-mounted camera. Video is recorded in multiple flight sections to capture civil structures in high resolution and is calibrated to correct lens distortion. Fiducial markers on the structure serve as virtual sensors for identifying marker and corner pixel coordinates. Homography transformation is used to determine marker position and orientation in real-world coordinates, followed by egomotion compensation to obtain absolute displacement and acceleration along with modal identification. The proposed method is validated through both laboratory and field experiments, demonstrating its effectiveness in structural modal identification. • This study proposes a contactless vision-driven modal identification method using a drone-mounted camera. • Fiducial markers and homography transformation are used to compute real-world positions, treating markers as virtual sensors. • Egomotion compensation is applied to extract the structure’s absolute response. • Modal parameters are identified using the SSI-COV technique. • The proposed method is validated through laboratory and field experiments.