所有產品在生產過程中自購進材料、製造加工到裝配成品，每一階段都需要檢驗，嚴格的品質管制可以確保產品品質。
本論文將致力於一套融噴非織物的瑕疵檢測分類系統，藉此系統來協助人工檢測。由於非織物的瑕疵檢測其處理方法不一致，故我們採用階層式的概念利用影像處理技術與最小距離法來辨識瑕疵，縮短檢驗的時間。在第一層計算灰階標準差來判定是否為瑕疵；於第二層先利用Log轉換調整對比和中值濾波消除雜訊後並使用統計式門檻值決定兩個最佳門檻值分割出瑕疵區域，接著計算白色區域灰階平均將破洞瑕疵辨識出來；於第三層利用標計化保留最大黑色區域並進行封閉運算使瑕疵輪廓更完整，接著計算緊緻性與SM值辨識聚合物滴液、併絲及摺紋三種非織物瑕疵。
由實驗結果得知，於第一層中輸入50筆學習樣本、47筆測試樣本，辨識率達100%；於第二層中輸入40筆學習樣本、39筆測試樣本，辨識率達100%；於第三層中輸入30筆學習樣本、24筆測試樣本，辨識率達100%，可明顯看出我們採用的階層式分類系統可以有效地應用在融噴非織物表面瑕疵的檢測工作上。

All of the products from purchasing materials and the manufacture to assembly the finished product in the productive process, must be inspected to the quality.
The thesis will devote to developing the surface defect inspection and classification system of the melt-blown non-woven fabric, and the system will assist artificial examination. We used the image processing technology and the minimum distance with the hierarchical concept to identify defects and reduce inspection time. The standard deviation of gray level is used to determine defect samples in the first stage. In the second stage, the log transformation is used to adjust contrast, the median filter is used to reduce the noise of images, and then I use the statistical threshold value decision method to choose two optimal threshold values for separating defect areas and fine the hole defects by the white region gray level mean. In the third stage, the labeling is used to retain the biggest black area and we use the closing operator to smooth the contour of defects. The compactness and second moment(SM) value are used as defect features to identify the three kinds of non-woven fabric defects, polymer shot, roping and fold.
From the experimental results, we have 50 studying samples and 47 testing samples in the first stage, the recognition rate is 100%. For 40 studying samples and 39 testing samples in the second stage, the recognition rate is 100%. When we have 30 studying samples and 24 testing samples in the third stage, the recognition rate is 100%. So we can take advantage of the hierarchical classification system to effectively inspect the surface defects of melt-blown non-woven fabrics.

[1] 余盛機編著，非織物製造技術，大學圖書出版社，1978。
[2] L. C. Sawyer and J. C. Chen, “Fiber Crimp Characterization by Image Analysis, Text”. U.S.A., Vol. 48, pp. 244-246, 1978.
[3] D. P. Thibodeaux and J. P. Evans, “Cotton Fiber Maturity by Image Analysis”, Text. Res. Journal, U.S.A., Vol. 56, pp. 130-139, 1986.
[4] X. C. Huang and R. R. Bresee, “Characterizing Nonwoven Web Structure Using Image Analysis Techniques”, INDA JNR. Vol. 5, No. 3, pp. 28-38.
[5] Z. Yan and R. R. Bresee, “Characterizing Nonwoven-Web Structure by Using Image-Analysis Techniques, Part V：Analysis of Shot in Meltblown Webs”, J. Text. Inst, Part 1, Vol. 89, No. 2, pp. 320-336, 1998.
[6] A. Cherkassky, “Evaluating Nonwoven Fabric Irregularity on Basis of Linnik Functionals”, Textile Res. J., Vol. 69, No. 10, pp. 701-708, 1999.
[7] T. J. Kang, S. H. Choi and S. M. Kin, “Automatic Recognition of Fabric Weave Patterns by Digital Image Analysis”, Textile Res. J. 69(2), pp. 77-83, 1999.
[8] J. J. Lin, “Applying a Co-occurrence Matrix to Automatic Inspection of Weaving Density for Woven Fabrics”, Textile Res. J. 72(6), pp. 486-490, 2002.
[9] C. F. J. Kuo, C. Y. Shih, C. Y. Kaoand and J. Y. Lee, “Automatic Recognition of Fabric Weave Patterns by Fuzzy C-Means Clustering Method”, Journal of Textile Research, (Accepted).
[10] 許大倫，數位影像處理技術量測在不織布棉網均勻度應用之研究，碩士論文，私立逢甲大學，2003。
[11] M. C. Hu and I. S. Tsai, “Fabric Inspection Based on Best Wavelet Packet Bases”, Textile Res. J. ,Vol. 70, No. 8, pp. 662-670, 2000.
[12] 林志豪，織品自動化檢測，碩士論文，私立義守大學，2006。
[13] 林東賦，應用影像處理技術與類神經理論於非織物瑕疵辨識，碩士論文，國立台灣科技大學，2001。
[14] 吳成柯等編譯，數位影像處理，儒林圖書有限公司SIM2538，1993。
[15] 余明興等編著，Borland C++ Builder 6 程式設計經典，文魁資訊股份有限公司，2007。
[16] R. C. Gonzalez, R. E. Woods, “Digital Image Processing 2nd ed.” by prientic-Hall, Inc 2002.
[17] D. H. Ballard and C. M. Brown, “Computer Vision”, Prentice-Hall, New York, 1982.
[18] N. Otsu, “A threshold selection method from gray level histogram”, IEEE Trans. On Systems, Man, and Cybernetics, Vol. SMC-8, pp. 62-66, 1978.
[19] R. Haralick, K. Shanmugram and I. Dinstein, “Textural features for image classification”, IEEE Transactions on Systems, Man, and Cybernetics, SMC-3, pp. 610-621, 1973.
[20] T. Matsuyama, K. Saburi, and M. Nagao, “A structural analyzer for regularly arranged textures”, Computer Graphis and Image Processing, Vol. 18, No. 3, pp. 259-278, 1982.
[21] A. E. Svolos and A. T. Pokropek, “An Evaluation of a Number of Techniques for Decreasing the Computational Complexity of Texture Feature Extraction through an Application to Ultrasonic Image Analysis”, Engineering in Medicine and Biology Society, 1997. Proceedings of the 19th Annual International Conference of the IEEE, Vol. 2, pp. 601-604, 1997.
[22] 鍾國亮編著，影像處理與電腦視覺第三版，東華書局，2005。
[23] L. H. Siew, M. R. Hodgson and E. J. Wood, “Texture Measures For Carpet Wear Assessment”, IEEE Transactions on Pattern Analysis and Machine Intelligence., Vol. 10, No. 1, pp. 92-105, 1988.
[24] 溫耀銘，應用小波封包轉換與類神經網路於針織物瑕疵辨識，碩士論文，國立台灣科技大學，2000。
[25] L. D. Stefano and A. Bulgarelli, “A Simple and Efficient Connected Components Labeling Algorithm”, Proceedinds of 10th International Conference on Image Analysis, pp. 322-327, 1999.
[26] M. E. Ruiz and P. Stinivasan, “Hierarchical Neural Networks for Text Categorization”, in proceedings of the 22nd ACM SIGIR International Conference on Information Retrieval, pp. 25-32. 1997.
[27] 沈育賢，道路交通標誌之偵測與辨識，碩士論文，國立台灣科技大學，2007。
[28] L. C. Wadsworth and S. R. Malkan, “A Review of Melt Blown Technology”, INB Nonwovens, pp. 2, 1991.