Computer vision based target-free 3D vibration displacement measurement of structures

Abstract

Vibration displacement response is widely used in the field of structural health monitoring (SHM) to monitor the health condition of civil engineering structures. Traditionally, the vibration displacement response is acquired by physical sensors, such as linear variable differential transformers (LVDT) and laser displacement sensors (LDS), or more commonly by accelerometers to measure accelerations instead of directly measuring the displacement due to the fact that direct displacement measurement requires a fixed platform to install the sensors. Recently, to overcome the difficulties in direct displacement measurement, computer vision based methods have become a research hotspot. This paper proposes an advanced binocular vision system for target-free full-field three-dimensional (3D) vibration displacement measurement of civil engineering structures. The state-of-the-art key point detection and matching algorithm based on deep learning is employed to achieve target-free measurement, which greatly improves the quantity and quality of matching natural key points with low contrast. The performance of the proposed vision based approach for 3D vibration displacement measurement is evaluated through an experimental test on a steel cantilever beam in the laboratory. The obtained vibration displacement responses obtained from the proposed approach are compared with those measured by LVDTs and LDSs. The results demonstrate that the displacement responses obtained from the proposed vision based approach are accurate compared with the traditional sensors, while the proposed approach is more cost effective and much easier to achieve accurate 3D vibration displacement measurement.