Design Of System For LCD Appearance Detects Detection And Recognition

With the diversiifcation of needs and applications, the industry of the liquid crystaldisplay develop rapidly in recent years. At the same time, people are increasingly highrequirements for liquid crystal display and don’t accept the product with poor quality. Inorder to ensure that products meet the user’s needs，many liquid crystal displaymanufacturers put in a lot of manpower and resources for product quality control. Thedetection eiffciency is not high by manual inspection in the past. The LCDs are oftenundetected and erroneous detected. The detection equipment based on machine visionbecome popular and most of them are mainly for the viewing area. There are fewequipments for the detection of the LCD panel.A system dedicated to the LCD screen appearance defect inspection andrecognition has been presented in this dissertation. It contains source illuminationsystem, eiffcient defect inspection and classiifcation method, simple and intuitivesoftware system.First of all, the modules of system of LCD screen appearance defects inspectionand recognition are introduced. As the high-quality image can effectively reduce thedifficulty of image processing, the study of light source is the key. Choosing the righttype of the light source and experimenting with different light source illumination angle.The purpose is to highlight the appearance of defects in the display and weaken thesurrounding area at the same time.Secondly, the focus on image processing algorithms in the system. In imagepreprocessing, binarization is an important step, selecting threshold value is the key ofbinarization. Contrasting sorts of threshold segmentation methods, the Otsu method ischoosed ifnally. Defect classiifcation and judgment is based on feature extraction.Rectangle and ellipse iftting has been experimented previous, but the accuracy is nothigh and ellipse iftting is relatively time-consuming. Improved rectangle algorithm hasbeen proposed and accurate judgment on the LCD screen appearance defects has beencome true.Thirdly, the appearance of defects are divided into four types through combining the demand for the actual testing and statistical analysis of a large number of defects.According to the processing of the four kinds of the defects products, the defects aresequenced by the seriousness. It is the basis for classiyfing defective products whennumbers of appearance defects appear on a LCD screen.Finally, the software system is designed speciifcally for inspecting the appearanceof the LCD screen. Software system veriyf the study of algorithms before and join thestandard control. It makes defect inspection to be flexible.Through lots of tests, the results show a very high accuracy of the system about thedefect inspection and recognition of the LCD screen. Effectively reducing undetectedand erroneous detected by manual inspection, and meet the needs of the productioninspection on line. At the same time, improved rectangle method based on improvedfeature extraction and classiifcation of the appearance defects in this dissertation havenot been reported in the literatures about LCD defect inspection in China.