Defects Detection of Printed Pattern on the Cylindrical

• Main Authors: LIN,JIA-CING, 林家慶
• Other Authors: Hsien-Huang Wu
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/91044738671020534620

碩士 === 國立雲林科技大學 === 電機工程系 === 102 === Currently, most screen printing of logo still conducted with manual inspection. Utilization eyes view all of marks is individually for obvious flaws then put into oven baking to complete the process of the screen printing. Ink is impossible to wipe after baking, so defect inspection of logo process need to very careful. Because the root of the daily production with thousands screen printing of sub-logo in the minimum allowable 1mm only. After a long time with manual inspection is easy to make eyes fatigue caused by misjudgment. In order to handle huge production and reduce false rate, this paper use Automated Optical Inspection (AOI) system hoping to replace or auxiliary requiring much time and effort manual inspection. In this paper, the experimental platform for desktop computers with screen printing machine, line scan camera and use on-Axis lights to capture images. Device under test (DUT) is cylinder which is logo, supposed to check the DUT mark whether there are flaws or not. The reason to choose line scan camera is that it possesses high frame speed and the effect of expand the cylinder's surface, which will be efficient to conduct image analysis. First, make the logo after imaging which is divided into several sub-logo, then save it as a standard template. Use normalized cross-correlation to alignment the various sub-logo in term to defect inspection. This paper using morphological to create two standard template and using logic operation whether there are flaws or not. After experimental results showed, in the case of cylinder is non-straight to generate by fewer deformation may be tolerated; Fiber, Void and Lose Print are out of tolerance (80μm), the logo can be detected. When detect a whole logo will take about 0.4 seconds there will be lose some information when accurate positioning, which is ignored some Contamination flaws.