Research On Mura Defect Detection Of LCD Screen Based On C-V Active Contour Model And Level Set

With the development of thin film transistor liquid crystal displays(TFT-LCD)toward high resolution,thinness,and low power consumption,the size of glass substrates and related optical components has gradually increased and the thickness has decreased,leading to various displays on TFT-LCD The probability of uneven defects(Mura)is greatly increased.Traditional artificial defect detection methods are seriously affected by subjective factors and the external environment.There is no uniform quantitative standard for defect levels,and it is difficult to meet the requirements of product quality and production efficiency.Therefore,it is necessary to design a set of Mura defect detection system that is fast,free from external environment interference,and meets human judgment standards.Mura,as a display defect with low contrast and irregular shape,has always been one of the most difficult to detect defects in LCD screens.Aiming at the characteristics of Mura defects,this paper proposes an improved C-V active contour model and level set method to achieve the segmentation of Mura defects under severe gray unevenness.The article builds a Mura defect detection system around screen area extraction and geometric correction,background texture suppression,brightness correction,and segmentation and quantification of Mura defects.The main research work and contributions of this paper are as follows:(1)As the captured images are affected by interference factors such as shooting conditions,background textures,and changes in lighting,this paper designs a pre-processing process suitable for LCD screen images.For the captured images that include backgrounds such as the console and situations where distortion may occur,the screen area after geometric correction is obtained through Canny’s algorithm,Hough line fitting,and perspective transformation.For the horizontal and vertical circuit textures contained in the screen area,16 sets of methods based on real-valued Gabor filter banks were designed to successfully remove the background texture.Regarding the brightness unevenness that may exist in the image,a certain correction effect is obtained by performing interpolation enhancement or attenuation processing on the comparison result of the grayscale average value in the sliding window and the global grayscale average value.(2)Aiming at the problem of poor segmentation of Mura defects by the CV active contour model under severe gray unevenness,this paper proposes a fusion of global information(CV)and local fitting information(Late)as fidelity The model of level set segmentation reflects the uneven distribution of gray in the image by Taylor expansion.At the same time,the energy penalty term and length term are introduced to avoid repeated initialization of the level set function,and to improve the speed and smoothness of curve evolution.Compared with the traditional active contour model,the improved active contour model successfully segmented the Mura defect in the case of severe uneven grayscale,and the number of iterations and segmentation of the model took less time.(3)A simple LCD screen Mura defect detection system was designed based on the on-site production conditions.It mainly includes camera capture part,defect detection,quantification part and inspection result display part.After field testing,the system’s defect false detection rate,missed detection rate,and detection time have basically reached the requirements of automated inspection.