以T^2統計量爲基礎之小波特徵多變量處理模式應用於表面瑕疵之檢測

A T^2 Statistics Based Wavelet Characteristic Multivariate Processing Model Applied to Automated Inspection of Surface Defects

10.29977/JCIIE.200403.0003

林宏達(Hong-Dar Lin)；陳志松(Chih-Sung Chen)

表面瑕疵檢測 ； Hotelling T^2多變量統計量 ； 小波轉換 ； 閥值切割技術 ； ripple-texture defect detection ； Hotelling T^2 multivariate statistics ； Wavelet transformation ； threshold techniques

工業工程學刊

21卷2期（2004 / 03 / 01）

121 - 135

繁體中文

本研究提出應用品質管制技術中Hotelling多變量管制圖之T^(2)統計量於表面瑕疵之檢測，發展小波特徵多變量處理模式(Wavelet Characteristic Multivariate Processing model,WCMP)，並以表層障蔽型半導性瓷片(SBL)上一種常見的水紋紋路瑕疵為檢測對象。而本研究則應用T^(2)統計量可整合多個影像特徵之特性，整合彩色影像之多個紋路特徵以降低水紋瑕疵檢測之誤判率，所提WCMP模式使用小波特徵作為影像之特徵值，並以T^(2)能量值之變化突顯異常瑕疵之存在，且後續搭配不同檢測之需要發展Mode-Double與Mode-Triple兩種閥值切割技術以確定瑕疵之位置。本研究提出之WCMP方法相較於其他紋路分析方法較具彈性，且有處理速度快之特性，而實驗結果也發現WCMP模式具有93.75％之SBL瓷片水紋瑕疵判斷檢測率，其搭配Mode-Double或Mode-Triple閥值切割方法切割水紋瑕疵，具有高達90％之水紋瑕疵位置偵測正確率。

Ripple-textures are common defects because of steam left on Surface Barrier Layer(SBL)chip surfaces.Ripple-texture defects influence not only appearances of SBLs,but also theelectronic properties of the products.The reasons why the inspection of ripple-texturedefects cannot be done automatically are:(1)the ripple-texture defect is semiopaque;(2)the ripple-texture is an unstructured texture;(3)edge gray levels of the ripple-texture defectare changed gradually.It is not only difficult to find the ripple-texture by artificialinspection,but also easy to make wrong judgments due to human subjectivity and eyefatigues.This study applies multivariate T^(2) statistics to automatic inspection of surfaceripple texture defects.We propose a WCMP(Wavelet Characteristic MultivariateProcessing)model,based on the Wavelet characteristics and multivariate processing,fordetecting the ripple texture defects.Modified threshold techniques are also presented forlocating the defects.Experimental results indicate that the proposed methods can detect93.75% images with ripple-texture defects and reach above 90% correct segmentation ofthe ripple-texture defect regions.

1. Alt, F. B.(1985).Multivariate quality control.Encyclopedia of Statistical Sciences.
2. Amet, A. L.,A. Ertuzun,A. Ercil(2002).An efficient method for texture defect detection: sub-band domain co-occurrence matrices.Computers and Geosciences,28,763-774.
3. Chan, C. H.,K. H. Pang(2000).Fabric defect detection by fourier analysis.IEEE Transactions on Industry Application,36,1267-1276.
4. Hotelling, H.(1947).Multivariate quality control.Techniques of Statistical Analysis
5. Jolayemi, J. K.(1995).A power function model for determining sample size for the operations of multivariate control chart.Computational statistic & data analysis,20,633-641.
6. Lowry, C. A.,D. C. Montgomery(1995).A review of multivariate control charts.IIE Transactions,27,800-810.
7. Lowry, C. A.,Woodall, W. H.(1992).A multivariate exponentially weighted moving average control chart.Technometrics,34,46-53.
8. Mallat, S. G.(1989).Multifrequency channel decompositions of image and wavelet models.IEEE Transactions on Acousyics Speech and Signal Processing,37(12),2091-2109.
9. Mallat, S. G.(1989).A theroy for mulitiresolution signal decomposition the wavelet representation.IEEE Transactions on Pattern Analysis and Machine Intelligence,11(7),674-693.
10. Martin, E. B.,A. J. Morris,C. Kiparissides(1999).Manufacturing performance enhancement through multivariate statistical process, control.Annual Review in Control,23,35-44.
11. Mason, R. L.,C. J. Youn(1999).Improving the sensitivity of the T^2 statistic in multivariate process control.Journal of Quality Technology,31(2),155-165.
12. Mason, R. L.,Y. M. Chou,J. C. Young(2001).Applying hotelling`s T^2 statistic to batch processes.Journal of Quality Technology,33(4),466-479.
13. Montgomery, D. C.(2000).Introduction to Statistical Quality Control.
14. Pgnatiello, J. J.,C. C. Runger(1990).Comparisons of multivariate CUSUM chart.Journal of Quality Technology,22(3),173-186.
15. Pikaz,A. Averbuch(1997).An efficient topological characterization of gray-levels textures using a multiresolution respresentation.Graphical Models and ImageProcessing,59,1-17.
16. Ramesh, J.,K. Rangachar,G. S. Brain(1995).Machine Vision.
17. Siew, L. H.,R. M. Hodgson,L. K. Wee(1988).Texture measures for carpet wear assessment.IEEE Transactions on Pattern Analysis and Machine Intelligence,10,92-150.
18. Tsai, D. M.,B. Hsiao(2001).Automatic surface inspection using wavelet reconstruction.Pattern Recognition,34(6),1285-1305.
19. Tsai, D. M.,C. Y. Hsieh(1999).Automated surface inspection for directional textures.Image and Vision Computing,18,49-62.
20. Tsai, D. M.,S. K. Wu(2000).Automated surface inspection using gabor filters.International Journal of Advanced Manufacturing Technology,16,474-482.
21. Tsai, D. M.,S. K. Wu,M. C. Chen(2001).Optimal gabor filter design for texture segmentation using stochastic optimization.Image and Vision Computing,19,299-316.
22. Wikstrnm C.(1998).Multivariate process and quality monitoring applied to an electroysis process. Part Ⅰ Process supervision with multivariate control chart.Chemometrics and Intelligent Laboratory System,42,211-231.
23. Wikstrom C.(1998).Multivariate process and quality monitoring applied to an electroysis process. Part Ⅱ-Process supervision with multivariate control chart.Chemometrics and Intelligent Laboratory System,42,233-240.

1. (2004)。PVA膠膜於曲面印刷之應用研究。中華印刷科技年報，2004，273-283。