Surface roughness classification for castings

摘要

In this paper we propose a machine vision system for the roughness classification of cast surfaces. Metal surfaces of castings in foundry processes appear as random and isotropic textures. The method of assessing surface quality is based on the two-dimensional Fourier transform of a cast surface in both gray-level image and binary image. The Bayes classifier and neural network classifier are implemented for roughness classification according to the discriminant features derived in the spatial-frequency domain. Experiments on cast specimens that contain nine roughness classes ranging from 6.3 to 400 μm have resulted in 100% classification accuracy rates.