應用機器視覺於增亮膜之表面瑕疵檢測

An Application of Machine Vision for Surface Defect Inspection of Brightness Enhancement Films

林巳軫(Ssu-Chen Lin)

指導教授 ： 林宏達 教授

朝陽科技大學/理工學院/工業工程與管理系/碩士(2009年)

摘要

摘要
背光模組(Backlight module)為液晶顯示器面板(LCD panel)的關鍵零組件之一，由於液晶本身不發光，背光模組之功能即在於供應充足之亮度與分佈均勻的光源，使其能正常顯示影像。在背光模組之零組件中以增亮膜（Brightness Enhancement Films,BEF）為材料成本較高之元件，其成本約佔背光模組材料成本比率約30~40%。且增亮膜的品質好壞亦會影響背光模組之光源分佈均勻性，有瑕疵之增亮膜讓背光模組無法提供充足之光源，將導致LCD 無法達到標準之輝度效果，故增亮膜之品質將影響背光模組之品質與成本。
增亮膜之表面瑕疵，目前仍以人員目視檢測為主，且鑑於該人員之經驗法則。此方式不僅檢測效率低，且無法達成固定之品質標準。本研究提出應用機器視覺系統檢測增亮膜之表面瑕疵。在瑕疵偵測技術方面，先使用二維傅立葉轉換及搭配累和管制法選取適當之刪除半徑，再利用巴特沃斯陷波濾波之方式將半徑內頻率值刪除並利用反傅立葉轉換技術將影像轉回空間域，最後以u+Kσ方式進行影像分割將瑕疵偵測出來。本研究共使用167 張影像之實驗結果顯示，本研究所提之方法在瑕疵偵測方面，平均檢出率為91.90%。

關鍵詞：增亮膜、瑕疵檢測、機器視覺系統、傅立葉轉換、累和管制法。

關鍵字

增亮膜 ； 瑕疵檢測 ； 累和管制法 ； 傅立葉轉換 ； 機器視覺系統

並列摘要

Abstract
The backlight module is one of key modules for LCD panel. Since LCD does not shine by itself, the function of backlight module is to supply adequate brightness and distribute uniform illumination, so that it could display regular images. In the backlight module, the material cost of the Brightness Enhancement Film (BEF) is about 30% to 40% which is much higher than other components. The quality of BEF also will affect the illumination of backlight module whether it is uniform. If a surface flaw appears in a backlight module so that it can not supply adequate light. This may causes the LCD could not achieve the desired illumination effects. Therefore, the quality of BEF will affect the quality and cost of backlight module.
The surface defects of BEF still are inspected by human’s eyes at present. According to inspector’s experience, this way is not only poor efficiency but also can’t reach to quality standard. This research applies machine vision techniques to inspect the surface defects of BEF. In our proposed approach, the Fourier transform and cumulative sum control algorithm are applied to choose adequate cutting radius in the Fourier domain first. And the Notch filter is used to remove the frequency values within the selected radius. Then, the inverse Fourier transform is conducted to transfer the filtered image back to the spatial domain. Finally, a simple thresholding method can be applied to separate the background and surface defects. Experimental results demonstrate that the surface defect detection rate of the proposed method is 91.90% which is higher than those of the tradition methods.

Keywords： Brightness Enhancement Films, Defect inspection, Machine Vision system, Fourier transform, Cumulative sum algorithm.

並列關鍵字

Cumulative sum algorithm ； Machine Vision ； Defect inspection ； Brightness Enhancement Filmssystem ； Fourier transform

參考文獻

[57] 黃朝義，「偏光膜介紹」，光電產業與技術情報，第20期，第40-47頁，台北(1999)。

[86] 張勝仁，「一個可增強車牌字元辨識效能的多重辨識器設計」，碩士論文，中原大學電子工程研究所，桃園(2003)。

[95] 林宏達、陳育洺，「見李以ANFIS模糊推論系統為基礎之電容瓷片外觀瑕疵分類與製程管制模式」，朝陽學報，第10期，第1-41頁，台中(2005)。

[1] Amet, A. L., Ertuzun, A. and Ercil, A., “Texture defect detection using subband domain co-occurrence matrices,” Proceedings of the IEEE Southwest Symposium on Image Analysis and Interpretation, pp205-210 (1998).

pp. 4251-4256 (1998).