Estimation of Metal Surface Roughness in Infrared Thermal Imaging Using Machine Learning

Abstract

In industries, surface roughness of metal play an important role for static and dynamic conditions. The surface roughness cause deformation and crack in additive manufactured components. Conventionally, the surface roughness measurement of metal is measured with profilometer. Although profilometer provides approximate results it lacks efficiency, mobility and requires a fixed space for measurement. In this paper, we propose infrared thermal imaging for surface roughness measurement of metal. Initially, the thermal image of metal is preprocessed to remove noise followed by thresholding and detecting porosity for metal grooves. Furthermore, Dyadic Wavelet Transform (DyWT) is applied for discontinuous metal groove edge detection. The obtained DyWT parameters are applied to the regression model to estimate the metal surface roughness. Finally, the surface roughness measurement of DyWT for 80, 150 and 400 mesh metal is compared with profilometer measurement. The DyWT measured surface roughness of metal was found to have 92% accurate.