目前對TFT-LCD產品的檢測方式依然是以人工檢測為主，針對IC線路是否對位與導電粒子是否足夠做檢測，這不僅會花費大量人力且判斷的標準也會隨著工作時間而有所變動，發展AOI自動檢測可以解決這些問題。而本篇論文介紹AOI檢測系統所需要的硬體與軟體，並針對其中的定位系統作分析，定位系統是用來執行TFT-LCD定位標記（Mark）檢測，實際測試常用的影像比對方式，並決定使用的影像比對方式，為了降低定位檢測時所花費的時間，使用階層式（Pyramid）來提昇檢測速度。IC壓合的位置也是重要的檢測項目，故針對位於定位標記上由IC壓合造成的壓痕執行偏移量檢測，比較常用的邊界偵測方式，並以速度與穩定性作為主。將影像比對的方式應用於Pad/Lead的搜尋上，自動搜尋並建立需要檢測的位置，降低建立檢測工單所花費的時間，讓使用者更容易運用檢測系統。

Manual inspection is the most popular way to perform TFT-LCD products inspection at the present time. Inspecting on whether the IC pad and panel are aligned and whether the particle amount on IC pad is enough are costing huge manpower. Also the accuracy of inspection decreases as the working time extends. Therefore we develop the AOI system to solve these problems. This paper describes the essential hardware and software of the AOI system and analyzes the alignment system within the AOI system. The alignment system functions as the inspection of TFT-LCD mark. We tested several conventional template matching methods and chose the most suitable one, also we use pyramid to decrease the execution time. Another important item is the IC press inspection. We inspect the position of IC press using bumping shift. Several frequently used edge detection methods are tested and one is chosen based on speed and reliability. We apply template matching on pad/lead searching, finding and creating position automatically to decrease the execution time to let the user manipulate the system more easily.

[1]E.H. Adelson, C.H. Anderson, J.R. Bergen, P.J. Burt and J.M. Ogden, ”Pyramid method in image processing”, RCA Eng., Vol. 29, No. 6, pp.33-41, 1984.
[2]A. Elmabrouk and A. Aggoun, “Edge-Detection Using Local Histogram Analysis”, ELECTRONICS LETTERS, Vol 34. Iss 12, pp 1216-1217, 1998.
[3]R.C. Gonzalez and R.E. Woods, Digital Image Processing (Second edition), Reading, Massachusetts: Addison-Wesley, 1992.
[4]Y.D. Li, “Automatic Defect Inspection and Enhancement for Patterned TFT-LCD Panel Surfaces”, Master thesis, National Cheng Kung University, 2006
[5]H.F. Ng, "Automatic thresholding for defect detection", in Proc. 3rd Int. Conf. Image and Graphics, Hong Kong, China, Dec.18-20, 2004, pp.532-535.
[6]W.M. Silver, “Normalized Correlation Search in Alignment, Gauging, and Inspection”, Proc. SPIE, Vol. 755, pp. 23–34, 1987.
[7]Y.T. Tasi, “Automatic Alignment and Inspection of TFT-LCD COG”, Master thesis, National Cheng Kung University, 2006
[8]S.D. Wei and S.H. Lai, “Robust and Efficient Image Alignment Based on Relative Gradient Matching”, IEEE TRANSACTIONS ON IMAGE PROCESSING, Vol.15, No.10, Oct. 2006.
[9]http://www.auo.com/auoDEV/technology
[10]http://www.chimei.com.tw/
[11]http://www.digital.idv.tw/DIGITAL
[12]http://www.digitimes.com.tw/