臺灣博碩士論文加值系統

本研究針對牛皮產業中皮革瑕疵，提供一快速皮革買賣自動檢測補償瑕疵面積的方法，以消弭皮革買賣面積計價供應間之紛爭。研究首先將各類瑕疵樣本依類別利用影像前處理、瑕疵區塊分析及根據皮革瑕疵面積、破洞情況、周長、周長比、瑕疵長度與寬度等六種瑕疵型態特徵擷取特徵值，來得到訓練之資料，然後再導入類神經網路上訓練，並將每一個運算元之權值存檔，來完成皮革瑕疵辨識之系統。然後再與扣點標準結合，模擬買賣狀況的面積差異情況，以達到自動辨識皮革瑕疵類別及補償瑕疵面積之目的。
研究結果顯示皮革瑕疵的辨識平均誤差率為3.34%，而考量扣點標準的誤差，其整合系統皮革瑕疵補償面積差異率為0.054％。驗證結果顯示本研究提供皮革買賣自動辨識皮革瑕疵及補償瑕疵面積一正確可行之方法。

Leather defects affect the usable area and the salable price. No international criterion specifies the compensatory counting for calf leather surface defects. So, complicated negotiation causes additional cost and argument between supplier and purchaser. This paper is to develop neural network learning method to implement the automatic recognition of types of leather defect and compensation of leather defective unusable area. We start by scanning the collected samples of defective leather and classifying the leather surface defects into seven types by the characteristic of the leather defective appearance. By using digital image processing technique, blobs analysis and the attributes of the area of leather defect, hole, perimeter, ratio of perimeter, length and width of defects etc., we can extract the characteristic of each type of leather defect and acquire the training data. Then by using the neural network learning technique, we can identify the types of leather defect automatically.
According to the compensatory leather defect unusable area standards corresponding to each type of leather defect. Practical leather recognition to transaction is simulated to evaluate the validity of the neural network technique. The results show that the mean error rate of recognizing the leather defect is 3.34% and integrate the compensatory leather defect standards, the mean deviation rate for the leather area of the simulated transaction is 0.05%. This provides an effective and reliable method for automatic recognition and compensation for finished leather transaction.

第一章 緒論………………………………………………………………1
1.1研究背景與動機…………………………………………………1
1.2研究方法與目的…………………………………………………2
1.3論文架構…………………………………………………………4
第二章 文獻回顧…………………………………………………………5
第三章 類神經網路訓練……………………………………………………10
3.1類神經網路理論…………………………………………………10
3.2倒傳遞類神經網路………………………………………………11
3.3類神經網路之訓練………………………………………………14
3.3.1 訓練樣本及網路之學習…………………………………15
3.3.2 設定網路學習參數………………………………………17
3.3.3 網路架構之設定…………………………………………18
第四章 皮革瑕疵自動辨識與補償面積系統……………………………19
4.1皮革瑕疵自動辨識及補償系統研究架構………………………19
4.2類神經網路訓練之架構…………………………………………23
4.3影像前處理………………………………………………………24
4.4瑕疵區塊分析……………………………………………………25
4.5皮革瑕疵特徵擷取………………………………………………32
4.3.1 瑕疵面積…………………………………………………34
4.3.2 瑕疵長度與寬度…………………………………………35
4.3.3 周長與周長比……………………………………………36
4.3.4 破洞………………………………………………………42
4.6輪廓分佈…………………………………………………………43
第五章 實驗辨識結果及分析……………………………………………46
5.1實驗設備…………………………………………………………46
5.2類神經網路訓練及辨識能力……………………………………46
5.3皮革瑕疵自動補償系統模擬……………………………………55
5.4皮革瑕疵分類之修正……………………………………………63
第六章 結論與建議………………………………………………………66
參考文獻…………………………………………………………………68
附錄一 影像標記程式……………………………………………………75
附錄二 實驗軟體操作說明………………………………………………80
附錄三 皮革瑕疵特徵值資料……………………………………………82
附錄四 隱藏層之權值……………………………………………………87
附錄五 皮革瑕疵辨識以及補償面積誤差模擬程式……………………94
附錄六 六類瑕疵類神經網路訓練之權數………………………………96
致謝
作者生平

1.皮革公會，台灣區皮革工業同業公會函文，台革軒字第二四四號，民國八十九年十月九日。
2.侯東旭與陳建諭，”使用類神經網路於工件辨認之研究”，亞太工業工程研討會八十三年論文集，第432-437頁(1994)。
3.洪蘭譯，天生嬰才：重新發現嬰兒的認知世界，遠流出版社(1996)。
4.彭德保、葉忠、陳麗真，”應用影像處理於皮革不可用面積之自動量測”，中華民國工業工程學刊，第17卷第6期第593頁至601頁，民國89年。
5.曹登發、盧炳勳，”類神經網路理論與應用”，全華科技圖書股份有限公司，民國81年。
6.葉忠、徐國政，”皮革供應建構瑕疵面積扣點標準之研究”，中區院校研究成果聯合發表會，成果專輯（一），B1-1，民國九十年六月。
7.衛祖賞，”數位影像處理”，全華科技圖書股份有限公司，第2至15頁(1988)。
8.葉怡成編著，”類神經網路模式應用與實作”，儒林圖書有限公司，民國87年。
9.蔡瑞煌，”類神經網路概論”，三民書局，民國84年。
10.蔡雅惠、蔡篤銘，”彩色圖形比對：元件蒐尋與瑕疵檢測之應用”，元智大學工業工程研究所碩士論文，民國87年。
11.蘇木春、張孝德編著，”機器學習 類神經網路、模糊系統以及基因演算法”，全華科技圖書股份有限公司，民國九十年七月。
12.繆紹綱 編著，數位影像處理─活用MATLAB，全華科技圖書股份有限公司 (2001)。
13.羅華強，”類神經網路-MATLAB的應用 (MATLAB 6.0)”，清蔚科技(2001)。
14.Alapuranen, P. and Westman, T.,”Automatic Visual Inspection of Wood Surfaces”, Proceedings of Eleventh International Conference on Pattern Recognition, August,pp.371-374(1992).
15.António F. Limas-Serafim, “Natural images’ segmentation for patterns’ recognition using edges’ pyramids and its application to the leather defects,” Proceedings IECON. IEEE, 1357-1360(1993).
16.António F. Limas-Serafim, “Multi-resolution pyramids for segmentation of natural images based on auto-regressive models: Application to calf leather classification,” Proceedings IECON. IEEE, 1842-1847(1991).
17.Attila Lerch and Dmitry Chetverikov “Knowledge-based Line-correction rules in a machine-vision system for the leather industry,” Engineering Applications of Artificial Intelligence, 4(6), 433-438(1991).
18.Branca A., M. Tafuri, G. Attolico and A. Distante, “Automated system for detection and classification of leather defects,” Optical Engineering, 35(12), 3485-3497(1996).
19.Chung Yeh and Der-Baau Perng, ”Establishing compensatory leather defect area standards”, submitted to the Journal of Robotics & Computer Integrated Manufacturing(2000).
20.Chung Yeh and Der-Baau Perng “Establishing demerit count reference for leather hides classification and grading standard,” The International Journal of Advanced Manufacturing Technology, 18:731-738(2001).
21.Dmitry Chetverikov and Attila Lerch “Prototype machine vision system for segmentation of hide images,” International Journal of Imaging Systems and Technology, 4,46-50(1992).
22.Gerard YAHIAOUI and Bertrand BOROCCO, “An Industrial Application of Neural Networks to Natural Textures Classification” International Workshop on Artifical Neural Network”, IWA NN’93, Sitges, Spain,1993.
23.Gonzalez R.C. and R.E. Woods., Digtal Image Processing, Addison Wesley(1992).
24.Gupta L., M.D.Srinath,“A Neural Network Approach to Robust Shpate Classification”, Internation Pattern Recognition, Vol.23, No.6,pp.563-568(1990).
25.Gupta L. and Srinath M.D.,“ Contour Sequence Moments for the Classification of Closed Planar Shapes”, pattern Recognition, Vol.20, No.3,pp.267-272(1987).
26.Hoang K. and A. Nachimuthu, “Image processing techniques for leather hide ranking on the footwear industry,” Machine Vision and Applications, 9, 119-29(1996).
27.Hoang K., W. Wen, A. Nachimuthu and X. L. Jiang, “Achieving automation in leather surface inspection,” Computers in Industry, 34, 43-54, (1997).
28.Horaud R., Olympieff S., and Charras J.P., “Shpate and Position Recognition of Mechanical Parts from Their Outlines”, The 1st Internation conference on Robot Vision and Sensory controls,pp125-134(1997).
29.Huynh L.V., “ A Vision System for In-Process Surface Quality Assessment,” SME vision 87,12-51(1987).
30.I-Shou Tsai, Chung-Hua Lin, and Jemg-Jong Lin,” Applying an artificial neural network to pattern recognition in fabbic defects.” Texible Res. J. 65(3), 123-130(1995).
31.Kittler J. and J. Illingworth. “Minimum error thresholding,” Pattern Recognition 19(l), 41-47, (1986).
32.Lovergine F.P., A. Branca, G. Attolico and A. Distante “Leather inspection by oriented texture analysis with a morphological approach,” Proceedings, International Conference on Image Processing IEEE Comput. Soc. Part 2, 669-671(1997).
33.Matorx Inspector V2.2, Matorx Electronic Systems Ltd. California, USA, (1997).
34.McCartney, J., Hinds, B.K. and Zhang, J. J. “LEATHER TEXTURE SYNTHSIS AND RENDERING”, Comput. & Graphics, Vol.18, No.1, pp. 87-92, 1994.
35.Moganti, M., Ercal, F.C., Dagli, H. and Tsunekawa, S., ”Automatic PCB Inspection Algorithms: A Survey”, Computer Vision and Image understanding, Vol.63, No.2,pp.287-313(1996).
36.Murino Vittorio, Ottonello Cinthya and Pagnan Sergio, “NOISY TEXTURE CLASSIFICATION: A HIGHER-ORDER STATISTICS APPORACH”, Pattern Recognition. Vol.31, No.4, pp.383-393,1998.
37.Otsu N., “A threshold selection method from gray-level histograms,” IEEE Trans. on Systems, Man and Cybernetics, Smc-9(1), 62-66(1979).
38.Parikh Jo. Ann, DaPonte John and Damodaran Meledath, “Texture Analysis of Ultrasonic Images Using Backpropagation Neural Networks”, International Workshop on Artifical Nerural Networks, 1993.
39.Pölzleitner W. and A. Niel, “Automatic inspection of leather surfaces,” SPIE 2347,50-58(1994).
40.Reeke. C.N., Sporns Jr. O., and Edelman G. M., ”Synthetic neural modeling: The”Darwin”” series of recognition automata. Proc. IEEE, Vol.78,pp.1498-1530(1990).
41.Rumelhart D.E., hinton G.E., and Williams R. J.,” Learning internal representations by error propagation,” in D.E. Rumelhart, J.L. McMcleland, et al., eds., Parallel Distributed Processing, Vol. 1,Chap. 8, Cambridge, MA:MIT Press(1986).
42.Schettini, R., “Multicolored Object Recognition and Location”, Pattern Recognition Letters, Vol.15, pp.901-933(1997).
43.Thomas C. Thorstensen, Practical Leather Technology, 4th, Edition, KRIEGER PUBLISHING COMPANY MALABAR, FLORIDA (1993).
44.Werbos P. J.,”Beyond regression: New tools for prediction and analysis in the behavioral sciences,”Doctoral dissertation, Harvard University,1974.
45.Yuen, P.C., Feng, G.C. and Tang,Y.Y.,”Printed Chinese Character Similarity Measurement Using Ring Projection and Distance Transform”, International Journal of Pattern Recognition and Artificial Intelligence, Vol.12, No.2, pp.209-221(1998).