A homography transformation based video stabilization method for structural health monitoring using a unmanned aerial vehicle

Using a computer vision technique to perform vibration-based structural health monitoring has advantages in that it is a low-cost non-contact approach, thus attracting a lot of attention. Great progress has been made recently. However, it is difficult for static cameras to be placed in an effective location in many scenarios, such as monitoring large-scale wind turbines, bridges, etc. The unmanned aerial vehicle (UAV) is a good choice to deal with this issue. In the past ten years, the application of UAVs in structural health monitoring was increasing dramatically. However, the non-stationary motion of a hovering UAV is a great challenge in terms of the estimation of dynamic characteristics. Thus, it is important to stabilise the video to achieve the level at which vibration can be monitored.
This paper puts forward a video stabilisation method based on homography transformation for the purpose of applying UAVs to vision-based vibration monitoring. Firstly, a motion model is built between the motion of the camera, called ego-motion, and its displacement on film. Secondly, an experiment is conducted to define the characteristics of a hovering UAV’s ego-motion. Thirdly, a stabilisation algorithm is developed for the compensation of ego-motion.