| Thin Film Transistor-Liquid Crystal Display(TFT-LCD)is occupying an increasing market share in the new display industry due to its small size,low power consumption and fast response speed.Competition among manufacturers It has also become particularly intense.Therefore,improving product quality and production efficiency has become an important factor for manufacturers to maintain their superiority.However,there are many types of defects in the TFT-LCD manufacturing process.The manual visual inspection method has disadvantages such as poor stability,time-consuming,and difficulty in uniform testing standards.However,the same detection algorithm cannot achieve effective detection of all defects under a single background.In this paper,the detection algorithm design and research are carried out for various defects in the TFT-LCD manufacturing process,and a set of TFT-LCD defect visual inspection system is designed.The preprocessing of the TFT-LCD display image is essential to the realization of TFT-LCD defect feature extraction,image segmentation,target detection and other technologies.This article has carried out research and design.Under the condition of ensuring that the surface of the display panel is free from dust and other noise interference,perform image averaging,Gaussian filtering,gamma transformation and histogram equalization processing on the collected display panel images to highlight important information in the image and change the quality of the captured image.Focusing on the characteristics and detection difficulties of different TFT-LCD defects,a real-time visual detection method for 3 types of defects under multiple backgrounds was researched and designed.Through the analysis of 6 different defects and detection algorithms,for line Mura,region Mura,and scratch defects,the largest inter-class variance method and morphological calculations are used to segment and perform defect image segmentation under 6 pure color backgrounds.ROI edge defect detection under the grid display background;for the smaller spot Mura,the Center Net network is used for defect detection under 6 kinds of pure color backgrounds;for the color deviation defect,it is judged by dividing the gray value variance and mean of the sub-region 32 The graded gradient color shows whether there is a color difference defect in the background.In view of the practical application of the later industry,through the analysis and research of the subject content and the TFT-LCD manual inspection process,a visual inspection device and system for display panel defects are built.At the same time,it also selects hardware devices such as industrial cameras,industrial lenses,and industrial display panels.Because the actual industrial scene is more complicated,it often leads to more noise in the TFT-LCD display panel image collected by the industrial camera.Therefore,this article designs a corresponding hardware solution to suppress noise.Moreover,the installation position of the camera in the industrial field is simulated,and an automatic adjustment device for the shooting height of the industrial camera is designed to adapt to display panels of different specifications.The research results of this paper show that the designed algorithm can better realize the real-time detection of TFT-LCD defects,and the detection effect is significantly improved and improved compared with traditional detection methods,which reduces the missed detection rate of TFT-LCD defects and has better performance.Applicability and practical value. |
PDF Full Text Request