Study On The Methods Of Machine Vision Inspection For The Mura Defect Of TFT-LCD

In recent years, with the development of TFT-LCD(Thin Film Transistor-Liquid Crystal Diaplay) towards large size, thin thickness,low power consumption and high resolution,the probability of Mura defects in the production process of TFT-LCD is greatly increasing. The traditional methods of relying on the human eye to detect Mura defect have been unable to meet the product quality and production efficiency of TFT-LCD. Therefore, the study on a machine vision detection method which is accurate, fast and without disturbance from outside has become a necessary requirement.A machine vision detection system with image acquisition module and image processing module is proposed to detect TFT-LCD Mura based on the analysis of Mura defect. The mainly responsibility of the image acquisition module is to capture and storage the images. While the function of image processing module, which includes four steps of image geometric correction, textural background suppression, Mura defect segmentation and quantification, is used to detect the Mura defect. According to the characteristics of Mura defect, three mainly key technologies of TFT-LCD textural background suppression based on the real Gabor wavelet filter, Mura defect segmentation based on Chan-Vese(C-V) model together with the level set method and Mura defect quantification based on DSEMU standard are solved by theoretical research and simulation. Firstly, according to the theory analysis of Gabor wavelet filtering principle and parameters selection, a real Gabor wavelet filter bank with four directions and four center frequency is designed. And a three levels sub image fusion method is also designed to suppress the image texture well. Secondly, aiming at the shortage of the traditional C-V model for the TFT-LCD Mura defect segmentation, two aspects at model self and approach are advanced to improve the segmentation capability and speed of the C-V model. Thirdly, a more accurate and reliable DSEMU standard is used to quantify the Mura defect because the traditional SEMU standard ignores the effect of the viewing distance on the defect quantification. At last a method to determine the position and shape of the Mura defect is given to improve the TFT-LCD production process and eliminate the Mura defect.Simulation results show that the machine vision detection method can successfully detect the Mura defect accurately and fast. And only two samples are not inspected accurately in 100 samples. The efficient detection rate is 98%, whereas the failure detection rate and miss detection rate is 2% and 0%, respectively. Therefore, the machine vision detection method can meet the production requirements of TFT-LCD.