Abstract
半導體工業是二十世紀的明星工業，特別是在臺灣。目前有越來越多的財團投資，以及許多優秀的人才投入其中。雖然在半導體的許多製程當中已經自動化，但仍然有部份檢驗工作需要靠檢驗員目檢，這樣做不僅浪費時間、而且檢驗員感覺視覺疲勞且負荷太重，常常因為檢驗員的個人主觀導致檢驗結果的不一致，使得品質控制變得困難。因此應用機器視覺技術發展自動化檢驗系統有其重要。這篇論文探討晶片內部銲線的自動化檢驗，在方法上包含兩部份，影像前處理以及晶片內部銲線瑕疵偵測演算法。針對銲線漏打及銲線斷裂這兩個瑕疵為研究重點，在檢驗方法發展完成後，配合不同的瑕疵型態和照明情況下進行一些實驗。結果顯示所提方法的辨識率可高達97% 以上，績效可媲美線上目檢人員，甚至在良好照明情況下，檢驗績效可勝過人。最後提出在真實的生產線上建立這套系統的方法。Semiconductor is a promising industry in the twenty century,especially in Taiwan, more and more financial groups invest inthis business. Although the manufacturing of semiconductorhas been automated in various processes,it still requires thehuman operators to perform the visual inspection task. Th taskis very time consuming, tedious,and fatiguing on the humaninspectors. Thus, the application of automated inspectionthrough machine vision has gained growing important. Thisthesis introduces the development of the post bondinginspection system in packaging process. The system includestwo main parts:(1)the basic image preprocessing techniques,(2)the PBI defects detection algorithm. This research focuseson the detection of missing bonding line and bonding linebreakage defects. Future,some experiments are conducted totest the system performance under different defect types andillumination levels. The results show that the performanceis the same as online inspectors, even better under goodillumination condition . The recognition rate is about 97%.Finally,discussion about online implementation of the systemare proposed.