Study On Key Technologies Of Non-destructive Visual Inspection For Quality Defects On Workpiece Surface

In order to realize the on-line inspection of the quality defects on machined surfaceof workpiece, it is a feasible solution to select non-destructive visual inspection based onmachine vision as the inspection method.Some key technologies in this inspectionmethod are studied in this thesis, including the construction of hardware and softwareplatform for image acquisition, the image texture analysis,the image de-noising,and theimage segmentation,et al.1.Hardware platform of image acquisition is constructed by series of products ofMicrovision, including placed platform of workpiece、industrial digital camera、imagedata acquisition card and illuminant,el at.The MVtec HALCON software is applied todevelop the programs for image acquisition.2.After comprehensive comparison, three kinds of multivariate-statistics-basedimage textural analysis approaches(principal component analysis,singular valuedecomposition,nonnegative matrix) are chose as potential means to suppress thedistraction caused by texture background for the inspection of quality defects. A textureimage from the data base of Brodatz is chose,and it is processed by the selectedapproaches.The feasibilities of selected approaches are verified.3.The selected image texture analysis methods are used to suppress the texturebackground of workpiece surface iamge, the curvature-based and theratio-threshold-based ideas are presented to determine the critical parameters of them.SSIM is adopted to assess the results of the suppression.4.First,Wiener filter is used to smooth the image noise which is introduced in thesuppression process;Then the defect regions are segmented and extracted by the methodof region growing; Finally,the noises of binary image are removed by the method ofmorphological.The results on processing the real images show that it is an effective way tosegment the quality defect regions from the texture background of workpiece surface bythe methods presented in this thesis.The study will has certain directing meaning towardthe realization of on-line inspection of quality defects on workpiece surface.