應用機器視覺方法於晶圓表面瑕疵檢測之研究

Applying Machine Vision Method in Wafer Surface Defect Inspection of Research

曾紀綱(Chi-Kang Tseng)

指導教授 ： 江行全

元智大學/工程學院/工業工程與管理學系/碩士(2004年)

摘要

由於現場Wafer檢測機台進行檢測時，易造成Type I Error誤判率過高，其檢測成功率約為50 %；因此本研究針對Wafer影像檢測發展一套Wafer表面瑕疵檢測系統，可使Defect突顯並確認瑕疵所在，以補檢測時Type I Error過高及減少後續人眼檢測人力。
本研究的檢測機制，首先是利用Gamma修正強化影像，將Wafer影像中較暗的地方加以突顯，屬於背景的地方給予濾除，再利用R、G、B Bands強化後的資訊，個別進行Maximum Between-Group二值分割，接著將R、G、B Bands強化分割後的資訊，利用本研究所提出的二值影像邏輯運算，將此三個Bands強化分割後的資訊加以合併，最後再將合併後的結果，利用精鍊式的Median Filter去除雜訊後，再判別待測樣本為一正常或瑕疵之Wafer影像。
利用本研究所提出的Wafer表面瑕疵檢測機制，針對廠商所提供的全部232個彩色Wafer影像樣本，其中人眼判定為正常影像有132個，判定為瑕疵影像有100個，將全部232個測試樣本以本研究的檢測機制進行檢測，檢測結果有24個測試樣本產生Type I Error，其Type I Error為18 %，及7個測試樣本產生Type II Error，其Type II Error為7 %，其整體之檢測成功率由原先50 %提升至87 %；檢測時間約在0.83秒~0.75秒內完成檢測。

關鍵字

晶圓 ； Gamma修正法 ； 二值影像邏輯運算 ； 瑕疵偵測

並列摘要

Because of high type I error rate resulted by using an optical inspection machine for a local wafer manufacturing company, the inspection successful rate is only about 50%. The focus of this research is to develop an inspection software that will significantly reduce the type I error rate and thus reduce the loading of the human operators.
The developed inspection method consists of firstly using the gamma correction method to enhance wafer images and to filter out the brighter background regions. Secondly, after abstracting the R、G、B bands information, the maximum between-group binary thresholding method was applied individually, and then the three enhanced information were combined through a set of logic operations on their binary images. Finally the noise of the image was reduced by applying an of iterative median filter.
Total of 232 samples were tested using the developed methodology. Among them, the operators identified 132 normal and 100 defect images. The results of using the developed method would results in 24 samples (18%) to produce type I error; and 7 samples (7%) to produce type II error. However, the whole inspection successful rate is from original 50% increased to 87%. The inspection time is above 0.75 to 0.83 seconds.

並列關鍵字

Wafer ； Gamma Correction ； Logic Operations Involving Binary Images ； Defect Detection

參考文獻

Bourgeat, P., Meriaudeau, F., Gorria, P. and Tobin, K. W. [2003]. “Content-based Segmentation of Patterned Wafers for Automatic Threshold Determination,” Machine Vision Applications of Industrial Inspection XI, Proceedings of the SPIE, Vol. 5011.

Chan, T. F. and Vase, L. A. [2001]. “Active Contours without Edges,” IEEE Transactions on Image Processing, Vol. 10, No. 2, PP. 266-277.

Chen, F.-L. and Liu, S.-F. [2000]. “A Neural-Network Approach to Recognize Defect Spatial Pattern in Semiconductor Fabrication,” IEEE Transactions on Semiconductor Manufacturing, Vol. 13, No. 3, PP. 366-373.

Cheng, S.-C. [2003]. “Region-Growing Approach to Color Segmentation Using 3-D Clustering and Relaxation Labelling,” IEEE Proc.-Vis. Image Signal Process, Vol. 150, No. 4, PP. 270-276.

Cho, K. D. and Tae, H. S. [2002]. “Improvement of Low Gray Scale Linearity using Multi-Luminance-Level Subfield Method in Plasma Display Panel,” IEEE Transactions on Consumer Electronics, Vol. 48, No. 3, PP. 377-381.