Manufacturing Technology Today

Open Access Open Access  Restricted Access Subscription Access
Open Access Open Access Open Access  Restricted Access Restricted Access Subscription Access

Obtaining subpixel level cutting tool displacements from video using a CNN architecture


Affiliations
1 Indian Institute of Technology Kanpur, Kanpur, India, India
     

   Subscribe/Renew Journal


To register motion from video of vibrating tools, acquisition must ensure that motion is spatially and temporally resolved. However, since tools often vibrate with subpixel level motion, and since cameras often trade speed for resolution, if acquisition is to respect the Nyquist limit to avoid temporal aliasing, then the spatial resolution is often not sufficient to detect small cutting tool motion. To address this problem, this paper shows for the first time that subpixel level tool motion can be inferred instead by using convolution neural networks. We train our model on a database using the phase-based optical flow scheme that is a subpixel level motion registration algorithm. Our model is shown to be capable of detecting small motion correctly. Though the frequency of vibration estimated from the registered motion is correct, further work is necessary on fine tuning model architecture to fix the errors observed in the estimation of damping.

Keywords

Convolution Neural Network, Vibrations, Phase-Based Optical Flow, Visual Vibrometry.
User
Subscription Login to verify subscription
Notifications
Font Size

  • Dosovitskiy, A., Fischer, P., Ilg, E., Hausser, P., Hazirbas, C., Golkov, V., Smagt, P. van der, Cremers, D., & Brox, T. (2015). Flownet: Learning optical flow with convolutional networks. In Proceedings of the IEEE International Conference on Computer Vision, 2758-2766. https://doi.org/10.1109/ iccv.2015.316

  • Fleet, D. J., & Jepson, A. D. (1990). Computation of component image velocity from local phase information. International Journal of Computer Vision, 5(1), 77-104. https://doi.org/10.1007/ BF00056772

  • Gupta, P., & Law, M. (2021). Evaluating tool point dynamics using smartphone-based visual vibrometry. Procedia CIRP, 101, 250-253. https:// doi.org/10.1016/j.procir.2020.09.196

  • Gupta, P., Law, M., & Mukhopadhyay, S. (2020). Evaluating tool point dynamics using outputonly modal analysis with mass-change methods. CIRP Journal of Manufacturing Science and Technology, 31, 251-264. https://doi. org/10.1016/j.cirpj.2020.06.001

  • Gupta, P., Rajput, H. S., & Law, M. (2021). Vision-based modal analysis of cutting tools. CIRP Journal of Manufacturing Science and Technology, 32, 91-107. https://doi.org/10.1016/j.cirpj.2020.11.012

  • Javh, J., Slavič, J., & Boltežar, M. (2017). The subpixel resolution of optical-flow-based modal analysis. Mechanical Systems and Signal Processing, 88, 89-99. https://doi.org/10.1016/j.ymssp.2016.11.009

  • Lambora, R., Law, M., & Mukhopadyay, S. (2022). Recovering cutting tool modal parameters from fractionally uncorrelated and potentially aliased signals. CIRP Journal of Manufacturing Science and Technology, 38, 414-426. https://doi. org/10.1016/j.cirpj.2022.05.014

  • Law, M., Gupta, P., & Mukhopadhyay, S. (2020). Modal analysis of machine tools using visual vibrometry and output-only methods. CIRP Annals, 69(1), 357-360. https://doi. org/10.1016/j.cirp.2020.04.043

  • Law, M., Lambora, R., & Mukhopadhyay, S. (2022). Modal parameter recovery from temporally aliased video recordings of cutting tools. CIRP Annals. https://doi.org/10.1016/j. cirp.2022.03.023

  • Luan, L., Zheng, J., Wang, M. L., Yang, Y., Rizzo, P., & Sun, H. (2021). Extracting full-field subpixel structural displacements from videos via deep learning. Journal of Sound and Vibration, 505, 116142.https://doi.org/10.1016/j.jsv.2021. 116142

  • Nuhman, P., Singh, A., Lambora, R., & Law, M. (2022). Methods to estimate subpixel level small motion from video of vibrating cutting tools. CIRP Journal of Manufacturing Science and Technology, 39, 175-184. https://doi. org/10.1016/j.cirpj.2022.08.005

  • Pinard, C. (2022). FlowNet Pytorch Implementation https://github.com/ClementPinard/

  • FlowNetPytorch, last accessed 2022/08/28.

  • Raizada, V., Rajput, H. S., & Law, M. (2023). Comparative analysis of image processing

  • schemes to register motion from video of vibrating cutting tools. Communicated for

  • consideration of presentation in the 11th CIRP Global Web Conference (CIRPe 2023) and for

  • appearing in the Procedia CIRP.

  • Rajput, H. S., & Law, M. (2023). Recovering cutting tool modal parameters from randomly

  • sampled signals using compressed sensing. Manufacturing Technology Today, 22(2), 17-22.


Abstract Views: 112

PDF Views: 0




  • Obtaining subpixel level cutting tool displacements from video using a CNN architecture

Abstract Views: 112  |  PDF Views: 0

Authors

Varun Raizada
Indian Institute of Technology Kanpur, Kanpur, India, India
Mohit Law
Indian Institute of Technology Kanpur, Kanpur, India, India

Abstract


To register motion from video of vibrating tools, acquisition must ensure that motion is spatially and temporally resolved. However, since tools often vibrate with subpixel level motion, and since cameras often trade speed for resolution, if acquisition is to respect the Nyquist limit to avoid temporal aliasing, then the spatial resolution is often not sufficient to detect small cutting tool motion. To address this problem, this paper shows for the first time that subpixel level tool motion can be inferred instead by using convolution neural networks. We train our model on a database using the phase-based optical flow scheme that is a subpixel level motion registration algorithm. Our model is shown to be capable of detecting small motion correctly. Though the frequency of vibration estimated from the registered motion is correct, further work is necessary on fine tuning model architecture to fix the errors observed in the estimation of damping.

Keywords


Convolution Neural Network, Vibrations, Phase-Based Optical Flow, Visual Vibrometry.

References