A evaluation of surface roughness classes by computer vision using wavelet transform in the frequency domain

This paper presents a multiresolution method based on the processing of surface images for unmanned visual quality inspection and surface roughness discrimination in turning. Sixteen machining tests were carried out using a constant depth of cut at 1.5 mm and different values for feed rate, in particular 0.15, 0.25, 0.4, and 0.5 mm/rev; cutting speed values were 250, 280, 320, and 350 m/min. With these cutting parameters, the roughness average values achieved covered a significant range between 0.8-14 mu m. The methodology proposed is based on the extraction of texture features from part surface images in the frequency domain using wavelet transform. In particular, one-level Haar wavelet transform is applied to the original surface images obtaining four sub-images: a smooth sub-image, a horizontal detail sub-image, a vertical detail sub-image, and a diagonal detail sub-image. These images are used for the extraction of features. Surface evaluation was accomplished by means of the analysis of gray levels in the vertical detail sub-image. Finally, a texture classification was performed by a multilayer Perceptron artificial neural network. Experimental results show that the proposed approach achieves error rates between 2.59% and 4.17%.