Liquid Crystal Display Screen Picture Quality Assessment System

With the development of computer science and technology,LCD displays have replaced CRT displays and become the largest product in the display area.The production of liquid crystal panels needs to be carried out in a dust-free environment and with precision technology.In spite of this,there are still various defects in the finished products.Traditional human eye defect detection methods are affected by human subjective factors,and there is no uniform evaluation standard for defect levels.Therefore,it is in line with human eye evaluation standards,and is highly efficient and stable.The research of the machine vision defect detection system is of great significance.A set of LCD screen quality evaluation system is designed.The whole system is mainly composed of three parts: the ARM microprocessor-controlled image acquisition module,the PC for defect detection and evaluation software.In the image acquisition module,the camera adopts a large area CCD camera.The cross movement mechanism assists the camera in anteroposterior,lateral,up and down directional movements,and the camera rotates axially to capture high-resolution,low-moiré interference source images,and the collected sources The image is transmitted to the PC via Ethernet;the defect detection module is mainly divided into image preprocessing;spot,plaque defect detection and line defect detection;Mura defect search in line with human visual characteristics;Mura defect based on active contour and level set methods Split,Mura defect quantification based on SEMI standard.Image preprocessing includes geometric correction,moir stripe removal,and uneven background illumination removal.The geometric correction part extracts the display screen of the liquid crystal screen by performing binarization,morphological opening operation and Hough straight line fitting to find the four corner points and the four-point perspective transformation processing.The moiré stripe is removed.In part,the moiré is effectively removed by axially rotating the camera and adaptive frequency domain filtering.The background is not evenly removed,and the background luminance of the image is adjusted dynamically based on the luminance equalization method so that the entire background is at approximately the same brightness.Point,spot,and line defect detection modules for pixel-level point defects with distinctly bright,easily detected features,compared with the direct gray-level half-value threshold segmentation and Laplace sharpening and half-value decimation The value threshold divides the principle and effect of segmentation of the two segmentation methods.It uses the Laplace sharpening and the half-valued threshold segmentation of the gray value,and counts the number of defect points and their position information.For the spot defect,the smoothing is designed.Image,background subtraction,binarization,connected-domain screening detection algorithms and statistical information on the number and location of plaque defects;for line defects,two smooth images,background subtraction,binarization,and Hough straight-line simulations were designed.The combined detection algorithm marks the line defect.For Mura defect low contrast,edge blur and other issues,in order to better fit the human eye's visual characteristics,based on the CSF function image enhancement processing,the processed image is defect segmented using active contour model and level set method,according to SEMI standard.With respect to the Mura defect rating method,the Mura defect is quantitatively evaluated by directly applying the paARMeters such as the defect contrast and area of preprocessed source image corresponding to the curve obtained during the segmentation.