Prediction of Surface Roughness by Machine Vision using Principal Components based Regression Analysis

Machine vision provides imaging based solution for inspection of quality. For components to perform their intended functions, surface quality is equally important as their dimensional quality. Surface Roughness (R-a) is a widely accepted measure for evaluating surface quality. Traditional methods for measuring surface roughness may not be feasible in the industries insisting on 100% inspection due to the efforts and time involved in measurements. Machine vision can provide an automated, economic, fast and reliable solution. This paper presents surface texture characterization of free hand grinding surfaces using machine vision approach and evaluation of their surface roughness using regression model based on machine vision parameters. Standard slip gauges sets of free hand grinding surfaces are selected for developing the regression model. Surface images are acquired in MATLAB software and processed for texture characterization using grey level co-occurrence (GLCM) matrix features. Then principal component analysis is carried out in SPSS software to define directions of unique variances in GLCM features. Furthermore, relationship between surface roughness value and GLCM features-based principal components is modeled using multiple regression analysis. Regression model developed can be used to predict unknown roughness values for free hand ground specimens. The approach demonstrated provides an automated, non-contact type method for measurement of surface roughness. (C) 2020 The Authors. Published by Elsevier B.V.