Wavelets and Artificial Neural Networks for Surface Roughness Estimation of Ceramic Tiles

Abstract

An important prerequisite for machined surfaces is the ability to estimate the surface roughness parameter. Non-contact methods are among the many measurement techniques that are frequently used to gauge the roughness of machined surfaces. These techniques are quick and adaptable. It is crucial in industries that produce ceramic tiles, glass, wood, and iron. This research proposes a novel method for measuring ceramic tile surfaces' surface roughness metrics through image processing. This system's acquired image is examined to see how its properties relate to those of the surface roughness. The energy details in terms of approximation, horizontal, vertical, and diagonal detail coefficients were derived after enhancement using a wavelet decomposition approach. The use of wavelet-based feature extraction in the evaluation of surface roughness was made justifiable by the energy details' strong correlation with the surface roughness parameter. Artificial neural networks (ANN) have been utilized to estimate Ra of the machined surfaces using the information collected from the wavelet transform of the pictures. Therefore, a correlation between image properties and Ra value has been attempted, effectively utilizing computer vision system for this application.