Research On The Detection Technology Of The Appearance Defect Of Polarizer Based On Structured Light

As an important part of the LCD panel, the aesthetic defects of polymeric polarizer have a serious impact on the quality of TFT-LCD panels. Currently, aesthetic defects of polarizer are mainly detected by manual work, however manual inspection is of subjectivity, and has low detection efficiency. AOI(Automatic Optic Inspection) system is widely used as an online detection system of polarizer, but the detection accuracy can not meet the requirements. So, the study of high precision and speed off-line automatic detection technology has significant impact on the development of related industries.Some slight and transparent defects of polarizer can hardly imaging and be characterized under conventional illumination. In order to solve this problem, the structured-light illumination was employed to enhance the defect’s contrast. So the subsequent image processing is the simpler and fast. Based on APG(Accelerated Proximal Gradient) algorithm and IALM(Inexact Augmented Lagrange Multiplier method) algorithm, RPCA algorithm was implemented and used to extract defects successfully. We also accelerate the classical APG algorithm to double speed, as a rusult, the accelerated APG algorithm reaches the same speed of IALM.Overexposure technology was used to detect the aesthetic defects of polymeric polarizer for the first time, that improves the defect’s contrast distinctly. As concerning, the histogram of the structured-light overexposure images’ characterized as two obvious peaks, three algorithms called as Column Means Threshold Method, Column Means Abnormal Waveform Inspection Method and Hole Filling Template Method respectively, were designed to identify the defects. It takes less than one second for these algorithms to extract the defect from an image of 2592*1942 pixels.The processing algorithms mentioned above were implemented on the platform of Lab VIEW and VDM(Vision Development Module) for defects recognition and features extraction. One hundred and four samples were tested. The inspection accuracies of the structured-light imaging with RPCA and the overexposure imaging with the three algorithms reach up to 100 and 99%, respectively. With a little lower accuracy, however the latter is more than 10 times faster than the former. The experimental results not only demonstrate this methodology’s validity to inspect the defects of transparent multilayer polymer films, but also reveal the potential of allowing real-time inspection.