Background texture suppression and defect extraction of high-speed milling surface image

Abstract

This paper proposed a method to extract the machine defect from the surface images of high-speed milling workpiece and reduced the impact of workpiece surface background texture. Firstly, Wiener filtering method is used to denoise the high-speed milling workpiece surface images. Then, the potential texture images are obtained by using the non-negative matrix decomposition algorithm for learning denoising images unsupervisedly. Next, original images and potential texture images are used to convolute with the imaginary part function of Gabor filter respectively and apply image differencing method to obtain the energy difference images. The binary defect images are extracted after the threshold segmentation method is applied. Application example shows the capability of this method to extract machine defect.