現行液晶顯示器產品製作完成後，全靠人工檢驗是否有缺陷，由於依靠人工檢驗測試，常因人員訓練不足或疲勞或其他種種因素，造成檢驗疏失。因未能及時檢驗出具有缺陷的產品，出貨到客戶端後將造成品質不良客訴及賠償問題，進而影響公司商譽。為了解決此問題，本論文結合加速強健特徵SURF (Speeded-Up Robust Features)演算法，提出用於液晶顯示器之自動圖像檢測方法。為模擬工廠作業方式，本論文建造一個暗室，並在內部架設一組網路攝影機，擷取液晶顯示器畫面的視覺影像。首先，將感興趣範圍的影像畫面，放大至液晶螢幕實際的解析度，再將其分割成九宮格區塊。其次，利用形態學之膨脹侵蝕濾波原理，濾除影像畫面中的雜訊，透過加速強健特徵演算法SURF找出圖像的特徵點。接著，將找出的特徵點，與載入的亮點樣本圖像進行配對，並將配對的亮點位置座標標示與紀錄。實驗結果顯示，本論文所提出的自動圖像檢測方法，可以快速又準確地辨識出液晶顯示器上壞點缺陷，有效攔截出不良品，以進行後製程處理。

In this thesis, we propose an automatic pattern inspection (API) method for liquid crystal displays (LCD) to replace the manual inspection. Manual defect inspection adopted in most manufactories nowadays is always affected by works’ status, such as lack of training, tiredness, or negligence. It may result in failure of inspection and send the defective products to customers. In order to solve this problem, an automatic pattern inspection method is proposed in this thesis. We establish a darkroom based on the real circumstance in the factory and utilize a webcam to capture the image of the LCD screen. The region of interest (ROI) is then enlarged until the image has the same resolution as the actual LCD. Then, the image are pre-processed by dilation and erosion to filter the noise. The Speeded-Up Robust Features (SURF) algorithm is integrated to calculate the feature points of the image and detect and compare with the bright points of sample image. The coordinates of the defect points are then marked and recorded for post processing. The experiments demonstrate the efficiency of the proposed API method and the bright points are recognized fast and successfully.

[1C. Harris and M. Stephens, “A combned corner and edge detector,” in Proc. Alvey Vision Conference, vol. 15, no. 50, pp. 10-5244, 1988.
[2]J. Shi, “Good features to track,” in Proc. IEEE Conference on Computer Vision and Pattern Recognition, CVPR’94, pp. 593-600, 1994.
[3]D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” International Journal of Computer Vision, vol. 60, no. 2, pp. 91-110, 2004.
[4]Logitech [Online]. Available: http://www.logitech.com/
[5]TOSHIBA [Online]. Available: http://www.grainew.com.tw/
[6]Microsoft Visual Studio 2010 [Online].
http://msdn.microsoft.com/zh-tw/vstudio
[7]H. Bay, A. Ess, T. Tuytelaars, and L. Van Gool, “Speeded-up robust features (SURF),” Computer Vision and Image Understanding, vol. 110, no. 3, pp. 346-359, 2008.
[8]J. Canny, “A computational approach to edge detection,” IEEE Transactions on Pattern Anaysis and Machine Intelligence, vol. 8, no. 6, pp. 679-698, 1986.
[9]M. Haralick, R. Sternberg, and X. Zhuang, “Image analysis using mathematical morphology,” IEEE Tansactions on Pattern Analysis and Machine Intelligence, vol. 9, no. 4, pp. 532-550, 1987.
[10]P. Viola and M. Jones, “Rapid object detection using a boosted cascade of simple features,” in Proc. 2001 IEEE Conference on Computer Vision and Pattern Recognition, CVPR’01, pp. I-511-I-518.
[11]A. Neubeck and L. Van Gool, “Efficient non-maximum suppression,” in Proc. IEEE 18th International Conference on Pattern Recognition, ICPR’06. vol.3, pp. 850-855, 2006.
[12]M. Calonder, V. Lepetit, C. Strecha, and P. Fua. “Brief: Binary robust independent elementary features,” in Proc. 2010 European Conference on Computer Vision, ECCV 2010, pp. 778-792.
[13]R. W. Hamming, “Error detecting and error correcting codes,” Bell Labs Technical Journal, vol. 29, no. 2, pp. 147-160, 1950.