透過您的圖書館登入
IP:3.133.79.70
  • 學位論文

應用支持向量機器於液晶顯示器面板瑕疵分類-以G公司為例

Apply Support Vector Machine in TFT-LCD Panel Classification - A Case Study on G Company

指導教授 : 張永佳 張桂琥

摘要


人們仰賴的資訊傳遞與溝通介面顯示器(Display)已經從過去極為流行的陰極射線管演進至現在的平面顯示器,其中液晶顯示器(Thin Film Transistor Liquid Crystal Display, TFT-LCD)為目前平面顯示器的主流技術。G公司為國內知名資通訊科技大廠,其品牌筆記型電腦是主力產品之一,而LCD即為該產品的顯示器技術。G公司欲按照供應商所分類之LCD Panel品質等級,組裝至不同規格的筆記型電腦上,而品質等級分類是依照瑕疵種類、數量以及位置而定。本研究欲應用自動光學檢測AOI(Automatic Optical Inspection),提高檢測速率,執行標準化之瑕疵判定,降低誤判或漏檢的情形,並達到最大化瑕疵分類正確率的目的。本研究將使用簡單的特徵向量快速判斷所有子影像是否有瑕疵,能先將大部分的正常子影像通過,接著再針對為數不多的瑕疵子影像進行辨認分類,如此能節省下不少處理時間。要進行各種特徵值的擷取與分析可以運用資料探勘技術來解決,且隨著近年來網際網路的發達以及資料庫技術的成熟,資料探勘的應用已越加廣泛且普遍,故本研究利用資料探勘中的分類技術「支持向量機器(Support Vector Machine, SVM)」來建立瑕疵分類模型,並改善檢測流程,執行符合G公司預期之LCD瑕疵分類。

並列摘要


Liquid Crystal Display (LCD) is the main technology of nowadays display for the applications of television, smartphone, laptop and so on. G company is a well-known enterprise for it’s branding laptop which screen is assembled by LCD and eager to match different level classification of LCD panel which depends on defect type, quantity and location to different specification of laptop. In order to do so, this study tries to apply Automatic Optical Inspection (AOI) to the inspection process to speed up the process of inspection and reduce the frequency of misjudgement. As a result, fast and precise inspection for the defects is a key issue for the rapid growth of the need of LCD nowadays. This study separates the model into two parts. In the first step, extracting the features of the LCD panel image to quickly determine whether the image is a defect one or not, thus, a big part of the normal images will pass through the process. Then focus on another small part of the remaining defect images by image processing to determine the type of defect. It can save a lot amount of time for only processing these small amount of images. The method of classification of this study is support vector machine (SVM) which is consider as an useful tool for data mining. As a result, this study applies support vector machine in TFT-LCD panel classification as the purpose of improving the process of inspection.

參考文獻


吳秋燕. (2007). 一維經驗模態分解法於 TFT-LCD 面板影像之 Mura (光源不均) 瑕疵檢測. 元智大學工業工程與管理學系學位論文, 1-150.
李政緯. (2011). 應用共變異矩陣表示法之加強型交通號誌偵測與辨識系統.
林思賢. (2006). 應用視覺及資料探勘技術於 TFT-LCD 陣列電路工程線中瑕疵辨識之研究-源/汲電極光罩瑕疵自動分類系統開發. 中原大學機械工程研究所學位論文, 1-122.
郭家成. (2006). 薄膜電晶體液晶顯示器 Mura 瑕疵檢測技術之研發. 臺北科技大學自動化科技研究所學位論文, 1-100.
陳志忠. (2001). 液晶顯示器的像素點缺陷與相對亮度均一性之自動化檢測.

延伸閱讀