Video-based vibration analysis of medium and high wood constructions

For vibration analysis of civil and large mechanical structures, video-based displacement measurement has emerged as an attractive technique due to its advantages such as non-contact, full-field measurement and low disturbance to occupants. Furthermore, there is a growing demand for timber buildings, as timber is seen as a sustainable alternative to concrete and steel. In this context, the thesis focuses on the development of a computer vision (CV)-based measurement protocol to measure the vibrations of high-rise timber buildings and large structures.

The initial phase of the study involved video-based displacement measurements using a camera network on a tower crane, chosen for its similarity in measurement conditions to timber high-rise buildings. The 3D displacement measurement methodology involves several key steps: (1) data acquisition using a camera network, (2) camera synchronization, (3) two-stage camera calibration (intrinsic and extrinsic), (4) motion tracking, and (5) structural 3D displacement calculation using non-linear triangulation. This approach provided full-field 3D displacement measurements that were validated with contact-based sensors such as accelerometers. In addition, operational modal analysis was performed on the tower crane using the 3D displacement data from the camera network, applying frequency domain decomposition to extract natural frequencies and corresponding mode shapes.

The thesis includes research at each step, including the development of a timestamp reconstruction method using a coded light signal to synchronize cameras without electronic triggering, a study on high-precision calibration for full-field displacement measurement, and the development of a line segment tracking method for robust motion tracking, complementing the Digital Image Correlation (DIC) method. These procedures will be replicated in the investigation of the dynamic properties of a full-scale mock-up of a timber structure using cameras.