摘要
由於液晶顯示器驅動IC小尺寸及表面複雜的圖樣特性，再加上潔淨度的要求，使得應用機器視覺之自動化檢測取代舊有的人工檢測方式已經勢在必行。因此，本文提出LCD驅動IC於料盤中自動檢測的方式取代現行的人工檢測，提高產品良率、產值及提升國內的檢測技術。由於承載LCD驅動IC之料盤具有製造公差，導致待檢測IC與標準樣板IC有旋轉與平移之變化，造成取像定位上的困難。因此，本論文先提出輪廓比對定位法，利用自行開發的物件標記演算法，將候選物件區分出來。接著，利用輪廓追蹤技術擷取物件之輪廓，並將物件之輪廓轉為形狀特徵，再利用最小距離分類器進行物件比對，完成輪廓比對定位。基於輪廓比對技術，本文提出相位移之旋轉角度估測技術來計算待檢測物件之旋轉角度。關於瑕疵檢測，有鑒於驅動IC在料盤上的定位誤差，本論文提出混合式之瑕疵檢測技術，一方面以可調式影像相減法來克服影像旋轉及平移的影響，進行複雜電路表面之瑕疵檢測；再結合物件標記與輪廓比對定位法檢測驅動IC上的凸塊，此混合式之瑕疵檢測技術可提升比對精度，並增加辨識的可靠度，使此檢測技術更能有效運用在LCD驅動IC之複雜圖像瑕疵檢測。
並列摘要
The requirement for high speed and high precision defect detection in LCD driver IC is rapidly growing because the pattern of circuit and gold bump on the IC surface has become more complex. Due to manufacturing tolerance of tray, the errors of image alignment from the driver IC positioning shift and rotation are always presented for in-tray inspection. Therefore, this thesis firstly compensates the positioning shift and rotation of in-tray IC using contour-based image alignment algorithm before defect detection. The contour-based image alignment algorithm consists of three steps: 1) proposed component-labeling, 2) shape descriptor transformation, and 3) phase-shifted orientation estimation. After applying the process of image alignment, the hybrid approach was applied for the defect detection of driver IC. This hybrid approach simultaneously detects the defects of the following two categories of surface: 1) the complicate surface, and 2) the primitive surface in the image. In this thesis, the modified image difference method is applied for the detection of complicate surface, and the design-rule strategy is adapted to detect the defects on bumps of LCD driver IC. Finally, the experimental results verify the proposed algorithms can accurately and rapidly inspect the defect for in-tray LCD driver IC. This algorithm is superior to the traditional template matching in defect detection, and the computational complexity can be efficiently reduced.