薄膜電晶體液晶顯示器(thin film transistor-liquid crystal display, TFT-LCD)的液晶胞製程(liquid crystal cell process)裡，包含一段磨邊程序，可將切割完後仍然不平整的薄膜電晶體基板(thin film transistor glass substrate)的玻璃邊緣磨平。在此程序中容易發生電晶體基板邊緣的破裂，進而破壞電晶體基板端子部(terminal)之電線結構，損壞液晶顯示器並降低生產品質。因此使用自動化的方式，檢測電晶體基板是否有再次磨邊的需求及磨邊後是否達到電晶體基板端子部品質要求，成為重要的議題。
本研究之檢查磨邊方面，首先使用模板比對(matching by correlation)的方式定位電晶體基板上的十字標示，並提出新的平均化模板(template)製作方法和門檻值決定方式，將定位精度提升到次像素定位的能力。本研究提出像素累加法和一維遮罩的概念，達到邊緣特質強化並去除雜訊等功效，接著以最大陡坡降法的觀念，定位出於電晶體基板邊緣之低對比度(contrast)的三層玻璃結構，且證明此方式具有優良的門檻值適用範圍。在端子部檢測方面，本研究應用模糊理論(fuzzy theory)和影像處理技術所開發之模糊運算子(fuzzy operator)，用以檢測端子部之瑕疵。其依據瑕疵具有弧形曲面之特徵，經由曲面分析和設計模糊規則等方式，將瑕疵分離出。本研究將影像金字塔的概念納入模糊運算子中，使模糊運算適用於檢測極端大小瑕疵之功能。經由實驗證實，模糊運算子於瑕疵檢測上具有2%之錯誤接受率，1.5%之錯誤拒絕率。結果以96.5%之總辨識率達到實驗目的之瑕疵檢測效果。
薄膜電晶體基板以本研究提出之檢查磨邊方法檢測，確實達成判別磨邊位置的需求，並使用模糊運算子對端子部檢測以確保端子部之完整性，達到自動化檢測的效果。

The final step of the thin film transistor-liquid crystal display (TFT-LCD) manufacturing process, where the TFT panel board undergoes the edge trimming process, may chap the edge of TFT panel board. The damage on the terminal department usually causes a short circuit, which will destruct the performance of the entire TFT panel board and decrease the product’s quality. Therefore, applying the automatic inspection system is crucial to this process, which examines the trimmed edge and protects the terminal department from harm.
In this research we used matching by correlation theory to locate the cross mark. In this theory we proposed a new method to construct an average pattern and a new threshold for selector function. These two approaches can increase the accuracy of the positioning. With edge inspection, we use the concept of maximum gradient descent method to locate the three glass layers on the TFT panel board. This concept additionally indicated that the threshold of this method was very robust. Regarding the inspection of the terminal department, we combined the fuzzy theory and image processing technology to create a fuzzy operator, which contains a 5x5 mask, the function of curve analysis, and the fuzzy system. The capability of this operator was adopted to examine the curve structure, which represents a defect since the terminal department should only contain straight lines. The fuzzy operator used the concept of the image pyramid theory, which allowed the fuzzy operator to examine the extreme size of defects. Our results indicated that the fuzzy operator on applied defect inspection had 2% of false acceptance rate, 1.5% of false rejection rate, and 96.5% of overall recognition rate.
The TFT panel board was further examined by proposed edge inspection system if the board needed to undergo the edge trimming process again. The fuzzy operator was again used in assurance of the quality of the terminal department.

[1] A. Goshtasby, S. H. Gage, and J. F. Bartholic, “A two-stage cross-correlation approach to template matching”, IEEE Transaction on Pattern Analysis and Machine Intelligence, vol. 6, no. 3, pp. 374-378 (1984)
[2] R. Brunelli and T. Poggio, “Face recognition: features versus templates”, IEEE Transaction on Pattern Analysis and Machine Intelligence, vol. 15, no. 10, pp. 1042-1052 (1993)
[3] D. M. Tsai and C. T. Lin, “Fast normalized cross correlation fordefect detection”, Pattern Recognition Letters, vol. 24, no. 15, pp. 2625-2631 (2003)
[4] D. M. Tsai, C. T. Lin, and J. F. Chen, “ The evaluation of normalized cross correlations for defect detection”, Pattern Recognition Letters, vol. 24, no. 15, pp. 2525-2535 (2003)
[5] J.P. Lewis, “Fast template matching”, Vision Interface, Canadian Image Processing and Pattern Recognition Society, Quebec City, Canada, pp. 120-123 (1995)
[6] I. Sobel and G. Feldman, “A 3x3 isotropic gradient operator for image processing”, presented at a talk at the stanford artificial project in 1968, unpublished but often cited, orig. in Pattern Classification and Scene Analysis , pp. 271-2 (1973)
[7] J. Canny, “A computational approach to edge detection”, IEEE Transaction on Pattern Analysis and Machine Intelligence, vol. PAMI-8, no. 6, pp. 679-698 (1986)
[8] C. W. Tao, W. E. Thompson, and J. S. Taur, “A fuzzy if then approach to edge detection”, in Proceedings Second IEEE Internationl Conference on Fuzzy Systems, FUZZ-IEEE 93’, San Francisco, CA, USA, pp. 1356-820 (1993)
[9] F. Russo “A new class of fuzzy operators for imageprocessing: design and implementation”, in Proceedings Second IEEE Internationl Conference on Fuzzy Systems, FUZZ-IEEE 93’, SanFrancisco, CA, USA, pp. 815-820 (1993)
[10] F. Russo and G. Ramponi, “Nonlinear fuzzy operators for image processing”, IEEE Transaction on Signal Processing, vol. 38, no. 3, pp. 429-440 (1994)
[11] F. Russo and G. Ramponi,“Edge extraction by FIRE operators”, in Proceedings Second IEEE Iinternationl Conference on Fuzzy Systems, FUZZ-IEEE 94’, SanFrancisco, CA, USA, pp. 249-253 (1994)
[12] C. H. Lee and L. H. Chen, “A fast search algorithm for vector quantization using mean pyramids of codewords”, IEEE Ttransaction on Cmmuniations, vol. 43, no. 234, pp. 1697-1702 (1995)
[13] C. Gonzalez and E. Woods, “Digital image processing third edition”, Pearson Education Taiwan Ltd, pp. 869-872 (2008)
[14] A. Kundu and J. Zhou, “Combination median filter”, IEEE Transactions on Image Processing, vol. 10, no. 3, pp. 433-429 (1992)