Visual vibration measurement of rotating shafts utilising optical flow enhanced by single-component image phase

Noncontact vibration-measurement methods based on machine vision have been widely applied for their high scene adaptability, low cost, and easy operation. Directly obtaining vibration signals of rotating shafts using machine vision is an innovative and efficient way to assess the vibration intensity and operational status of rotating mechanical systems. However, their measurement accuracy is limited by lighting, noise, image resolution, and other factors. This work introduces a visual vibration-measurement method utilising optical flow enhanced by single-component image phase (SCIPOF). This method separates single-component images from the original images by designing a narrowband bandpass image filter with highly concentrated frequency components. Subsequently, analytic signals are constructed using these single-component images and their first order derivatives, from which the phase of the images is extracted. Finally, the phase variations from the reference image to other images in the videos are converted to the actual displacement signal of shafts. The proposed method avoids the effect of manual parameter selection of existing methods on measurement accuracy, which improves its robustness. A series of numerical simulations and actual rotor experiments demonstrate that SCIPOF outperforms existing methods in accuracy and reliability, and maintains high accuracy even with low signal-to-noise ratio video data.