| Summary: | 碩士 === 國立臺灣科技大學 === 高分子系 === 95 === This research applied digital image processing technology to the defect inspection of polarizer products. It first used line scan charged couple device (CCD) to screen the defect images, and then reduced the defect image ( pixel) to four times smaller using downsampling compression method, and then conducted the Wavelet transform (WT) on the four times smaller image for two times to reduce the size to 16 times smaller, and then the image size was pixel, so that it became the right size for the following image processing. Then the power-law transform was used to intensify the characteristics of the defect, and then processed it with lower-upper-middle (LUM) filter which could delete noise background while reserve the edge characteristics of the defect. And then the edge of the defect would be found out through the screening of the isotropic results after 45 degrees and 90 degrees increment of Laplace operator, and then used the statistic threshold method to separate the defect image, and in the end used Hough transform to distinguish the spot defect as dust, foreign matter and beat print, air bubble, and the line defect as scratch mark. This research picked out 200 sheets of polarizer defect images, then it investigated them by self-developed inspection system, the experimental result showed that it could test all the flaws on the 200 sheets of images, and successfully distinguished the spot defect and line defect, the distinguish rate is 100%, and it only cost 0.9 sec to detect a sheet. Therefore it has clear that this research has successfully developed a set of on-line real time defect inspection system applicable to the polarizer.
The polarizer automatic optical inspection (AOI) system developed by this research could overcome the inconveniences and misjudgments caused by the manual inspection, and it would inform the producers whether the defect on the polarizer was spot defect formed in the manufacture or line defect in the following transportation, so that the producers could make different improvements for different types of defects.
|
|---|
