Research On Machine Vision Inpection System Of OLED Display

OLED(Organic Light Emitting Diode)screen has a wide range of applications because of its advantages in power consumption and angle of view.However,in practical applications,OLED screen often has three main defects,including point defects,line defects and MURA defects.Among them,the size of tiny point defects is micrometer level,which brings large difficulty to artificial eye detection.MURA defects generally have the characteristics of low contrast,blurred edges,unfixed shape and difficult judgment of the naked eye,those reasons caused it is the most difficult part to detect defects.In addition,OLED screen also needs color detection and character sets detection.Character sets detection usually outputs all character sets(ASCII sets)to the screen,so that the naked eye need judge dozens of characters.Such a detection method takes a long time and the reliability is also difficult to be guaranteed.In view of the above mentioned problems,this paper adopts machine vision instead of manual detection,and it also solves the problem of MURA defect background interference that can't handle with mainstream machine vision software.The hardware platform uses industrial cameras and universal X86 servers.In software,we use the Ubuntu operation system and Open CV image processing platform to reduce the cost and the difficulty for the software development of test system.Aiming at the most critical part of illumination in machine vision,this paper designs a good shade tool based on the self-luminous property of OLED screen,so that we can get the feature value of defect very clearly.The background interference is the main difficulty of MURA defects detecting,that is,the weak eigenvalue of the MURA defect will be submerged in the periodic background interference.In this paper,the image correction algorithm is used to correct the distorted image,and then the periodic background interference is suppressed.Due to the characteristics of filter usually will destroy the MURA defect value,through the comparison of various filtering algorithm,Gabor filtering and SVD(Singular Value Decomposition)has the ability of suppressing periodic background interference.In this paper,we process the corrected image by combining the SVD and Gabor filtering method.The periodic background interference is filtered under the premise of retaining the MURA defect feature value,and then the difference operation is performed with the no defect template image.We can detect the point,line and MURA defects by judging the result of the difference.For color aspects,we will transform the RGB(Red,Green,Black)color space into HSV(Hue,Saturation,Value)color space,then judging whether the color meets the requirements by H value.The test of character sets can detect the character quickly by dividing the characters in the specific area and using the template matching.Finally,the detection system and detection algorithms are validated by using nearly 100 OLED screen samples which contain all the above types of defects.