Research Of Key Technologies For TFT-LCD Display Defects Detection System Base On Machine Vision

With hundreds of processes,the process of manufacturing the LCD screen is extremely complex,there are inevitably kinds of defects.So detecting of defects in LCD production process plays a key role.Using machine vision technology to achieve high speed automatic detection of LCD light defect is one of the important research topic in the LCD automatic production.In this paper,through the analysis of the characteristics of defects and flaws in the image of LCD screen lighting,we proposed fast and effective method of defect detection based on image processing.This paper mainly researched on the following aspects:Research on sub-pixel defect detection: the traditional method with image processing is difficult to detect the sub-pixel defects.We analyzed characteristics of LCD image,and found that the arrangement of pixels is in order and same for the same type of LCD screen,so spectrum of different images is the same through Fourier transform of images.For this,we proposed the method of sub-pixel defect detection based notch filtering and image registration.Firstly,we established registration template and notch filtering template with no defect template;and the image registration is finished with registration template;then we use notch filtering template for filtering processing,so that background texture can be eliminated,which made the defect more obvious.Finally,we accomplish threshold segmentation,and found defects.The results showed that the method can accurately and quickly detect sub-pixel level defects.Research on small defect detection methods: since a narrow range of gray contain various information,such as defects,background,details,noise,as a result that it is difficult to distinguish information.Therefore,we proposed the LCD screen defect detection methods.We combined local entropy of pixel distribution state and local uniformity of that,in order to eliminated noise within the neighborhood space and obtain the distribution of pixel space,which we accomplished the detection of Medium and small defects with spatial characteristics.Through simulation analysis,this method improves the speed of detection,and accurately detects defects without frequency domain processing,what's more,it has better robustness than other methods.Research on Mura defect detection method: Mura defect detection of LCD screen has the overall background defects of uneven brightness,and the change of grayscale is not significant,the defect detection method based on machine vision is very difficult to detect.We proposed a new Mura defect detection method,firstly,the difference between the background of LCD screen image and gray value of Mura defect was analyzed,and then the background clutter was suppressed through the mean filter and background subtraction method,we afterward used the gray constraint to get the suspected defect area Mura.Finally,gray features of suspected areas were put into the trained BP neural network for extracting defect target.The results showed that this method has better effect on eliminating noise,and the correct rate of detection is higher.