Non-contact surface roughness measurement of blasted specimens using machine vision technique

This article describes non-contact surface roughness measurement for shot and grit blasted specimens. A simple machine vision algorithm was proposed for determining ten-point mean roughness (Rz) of a specimen after shot and grit blasting. Surfaces cleaned with different blast abrasives were captured by a microscope. Binary images were then created by using Otsu's thresholding method. On the binary image, bright regions were selected, associated with specular reflection of light. The average size of selected regions was calculated in this proposed method. It was found that the average size has a linear relation with ten-point mean roughness of both shot and grit blasted surfaces. It was also identified that the parameter of a linear regression for shot blasting is different from that for grit blasting. On the basis of the relation, it is possible to determine the ten-point mean roughness of a blasted surface with a captured surface image.