Automatic Optic Inspection For Defects Of ITO Patterns In Capacitive Touch Panel

Capacitive touch panels(CTPs), as a medium of information interactions, have become essential parts in many consumer electronics such as tablets, smart phones and ATM, etc. In CTP manufacturing, the important procedure is to etch ITO patterns on the substrates, which have defects or not will determine the quality of the CTP. Thus, defect detection of etched ITO patterns on the substrates is crucial. In this thesis, we research on the key technologies related about automated optical defect detection for it. The main contents of this thesis are:(1). Design of the high-contrast imaging system based on max superimposed reflection coefficient. The ITO patterns are near transparent under the visible light so that it cannot be effectively differentiated from substrates in images. To address this problem, we establish a mathematical model for the imaging system first. And then, considering cost and hardware, we construct a high-contrast imaging system based on max superimposed reflection coefficient, which can make discrimination of the two materials to be 15.5%.(2). Autofocusing system based on Brenner value of 1-D line images. Installation errors and lack of experiences will lead to the problem of setting improper focusing. This problem will definitely lead to unclear imaging with manufacturing errors. For solving this problem, we establish a model for the coaxial illumination system so as to find the formulas of depth of field. And then, according to the Brenner values of 1-D line images to realize autofocusing. Experimental results show this system can neutralize those errors and automatically find the best focusing plane.(3). Nonnegative Matrix Factorization(NMF)-based image registration method. The image of ITO patterns are very large due to the high resolutions of line-scan CCD, we proposed a NMF-based image registration method mainly for low efficiency of big image registration. The experimental results show the method can accurately register two images, eliminate the computation of registration and promote the efficiency.(4). Defect extraction methods. In this thesis, we researched on defect extraction methods such as 1-D Image Comparison Algorithm, NMF with tolerance model and notch-filtering template. And the methods are analyzed their advantages of application. Moreover, the methods are parameterized design so that it can automatically adjust for different kinds of ITO patterns.(5). Defect classification method based on the changing times of defect boundary. Based on the characteristics of ITO patterns in the images, we proposed a defect classification method based on the changing times of defect boundary. This method is combined with the defect extraction methods of 1-D Image Comparison Algorithm, NMF with tolerance model. Experimental results show this method is accurate, fast and it can solving the classification of various defects.