The Texture Background Suppression And Defect Extraction Of Milling Workpiece Surface Image

The defect detection of milling workpiece surface can be achieved by using the machine vision technology quickly, but the texture background of milling workpiece surface always lead to the failure of the defect detection in the practical application. In order to analyze and extract the manufacturing defects in the texture background of workpiece surface image quickly and effectively, a surface defect extraction method based on NMF and the imaginary part features of Gabor wavelet is proposed in this dissertation. The main research work is as follows:1. The high-speed milling experiments. A series of high-speed milling experiments by using high-speed double turntable five-axis linkage machining center(VMC- C30) as experimental platform were done, the experiment contents include choosing machining tools 、 process parameters and the type of cooling fluid, determining the machining experiment scheme of particular workpiece material. The roughness tester is used to detect the quality of workpiece surface after milling.2. To obtain the surface images of milling workpiece. The hardware platform of workpiece surface image acquisition is constructed by series of products of Microvision, including placed platform of workpiece、industrial digital camera、image data acquisition card and illuminant, el at. At the same time, the workpiece surface images are collected by calling the camera application of the software platform which is constructed by MVIPS software.3. The defects identification of milling workpiece surface. The wiener filtering method is used to denoise the high-speed milling workpiece surface images. The potential texture image is obtained by using the nonnegative matrix decomposition algorithm for learning denoising images unsupervisedly. Then the original image and the potential texture image is used to convolute with the imaginary part function of Gabor wavelet filter respectively, and the difference method principle is used to obtain the energy distribution difference between two images. The defect features of workpiece surface denoising images are extracted after the energy distribution difference threshold is determined. According to the test results of workpiece surface quality, the geometrical features of the high-speed milling workpiece surface defects can be identified after the analysis of defect features. It means that the digital description of defects is implemented.The high-speed milling experiments and the results of image processing in this dissertation showed that the surface defect extraction method based on NMF and the imaginary part features of Gabor wavelet can restrain the background texture effectively, improve the contrast between background texture and defects, achieve the digital description of the workpiece surface defects, and lay a solid theoretical foundation for using machine vision technology to detect the workpiece surface quality better.