Comparative Analysis of Image Processing Schemes to Register Motion from Video of Vibrating Cutting Tools

This paper reports on the efficacy of six different image processing schemes to properly resolve cutting tool motion from video. We compare the edge detection and tracking scheme with intensity- and phase-based optical flow-based schemes and with digital image correlation. We also introduce the use of two new schemes: one based on particle image velocimetry (PIV), and another based on the use of convolution neural networks (CNN). Methods are illustrated with video of two tools. Results are benchmarked with twice integrated accelerations. We find that the edge detection and tracking scheme, the PIV scheme, and the CNN scheme are robust.