晶粒表面缺陷自動視覺檢測系統之設計與開發

Abstract

半導體製造業對產品品質的要求相當嚴謹，因此在晶粒(Die)封裝前的晶粒缺陷檢測是一個品管很重要的過程，傳統的晶粒表面缺陷檢測通常是目視檢測，此方式需要花費大量的人力使用肉眼判斷，不同檢測人員對缺陷的判斷可能不一致，且容易因視覺疲勞產生錯誤判斷或是判斷標準不一致的情形發生。 本論文擬針對晶粒表面需要檢驗的缺陷部分，包括：(1)微粒或污染、(2)面積缺損、(3)變色與(4)護層不良等，利用機器視覺技術，提出了一個自動化的檢測演算法。 本論文之研究目的在於發展一套以自動視覺檢測系統為基礎的晶粒表面缺陷檢測系統，能檢測出晶粒表面中具有缺陷的晶粒。以提升檢測效率、減少成本，並達成可進行全檢的三項目標。

Product quality is an important factor in semiconductor manufacturing. Therefore, die defect inspection is an important quality control process before packaging. Conventionally, the inspection of die surface defects by human observation is labor intensive. It results in low efficiency and inaccuracy. This research is to design and develop an automated visual inspection system for die surface defects by using the machine vision technology. The mainly focused inspection items of dice are particles, contaminations, pad missing, pad damage, discoloration, and passivation. A prototype of the automated visual inspection system for die surface defect inspection will be implemented for inspection efficiency, cost down, and full-inspection.