Estimation of Surface roughness parameter using Wavelets based feature extraction

Practical use of Computer Vision for surface roughness estimation faces many challenges, as only image is used for evaluation rather than the components. Existing schemes are less suited for real-time machine vision applications due to the great computational burden involved in processing a large image. For example, an operation such as rotation and scaling involves four multiplications and four additions per pixel, which is going to be computationally complex. In this paper, the quantitative measure of surface roughness is estimated in the spatial frequency domain using wavelet transformation (WT) by extracting four features namely total energy (E-t), energy horizontal (E-h), energy vertical (E-v) and energy diagonal (E-d). An exhaustive analysis is done with comparison studies wherever required to make sure that the proposed method of estimating surface finish based on the computer vision processing of image is more consistent. The predicted surface finish values using WT are found to correlate well with the conventional stylus surface finish (R-t) values.