Research On Key Techniques Of ROI System For LC Cell Defects Of Small-Size LCD

The degree of automation of the manufacture of LCD is very high, however, thedefects detection rely on manual inspection. The defects detection is very important forcontrolling the product quality, search the reason and feedback in manufacturing process.The objective of this thesis is to delibrate a machine vision-based automated opticalinspection equipment, focuses on small-size LCD cell defects detection, including spot,scratch, spacer gather, color edge and LC leakage. The research of automatic inspectionequipment will promote quality control and standard detection of LCD.The contents of this report are as follow.1. A optical experiment platform have been built including backlight, polarizers,camera, by research the causes and phenomena of LCD cell defects. And high-qualityimages have been captured.2. A method for extracting regions of interesting is proposed. The method extractsthe contour pixels of the LCD in the image, and according to the rectangularcharacteristic of the liquid crystal panel, a rectangular ROI fitting is presented, and theregion containing the image of the target area only is obtained. The method hasimproved the efficiency and reduce the difficulty of detection.3.The causes and characteristics of the LCD texture is analysed, and a method ofeliminating texture is proposed, which extracting the standard LCD texture to eliminatethe texture in the image of the LCD that to be detected, false detection rate has beenreduced and detection accuracy has been improved.4. The defects are divided into two parts as brightness defects and chroma defectson the basis of defects phenomena. The method that the LCD image is converted fromthe RGB color space to the HSI color space is discussed, in order to obtain a highcontrast defects image.5. The double-threshold segmentation and edge-based segmentation is described bythe characteristics of LC cell defects in HSI color space, and experiment shows thatthese methods have good segmentation quality to all defects.6. The characteristic value extraction according to regional central moment is presented, the method can accurately estimate the defect area, length, width, angle andflat degrees.