研究生: |
許裕騄 Hsu, Yu-Lu |
---|---|
論文名稱: |
棒鋼表面缺陷自動化檢測系統 A Computer Vision System for Automatic Steel Surface Inspection |
指導教授: |
孫永年
Sun, Yung-Nien |
學位類別: |
碩士 Master |
系所名稱: |
電機資訊學院 - 資訊工程學系 Department of Computer Science and Information Engineering |
論文出版年: | 2007 |
畢業學年度: | 95 |
語文別: | 中文 |
論文頁數: | 80 |
中文關鍵詞: | 相關向量機 、倒傳遞類神經網路 、電腦視覺 、缺陷偵測 、缺陷分類 |
外文關鍵詞: | Back-Propagation Neural Network, Relevance Vector Machine, defect classification, defect detection, computer vision |
相關次數: | 點閱:131 下載:5 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
本篇論文提出一套自動化對棒鋼表面缺陷進行檢測的系統,此系統分析棒鋼影像以偵測其是否存在缺陷,並對缺陷進行分類,整個系統可以分為三個模組:缺陷分割模組、特徵擷取模組和缺陷分類模組。缺陷分割模組的目的是要明確得知缺陷的位置,並將其框選出來,其包括了影像前處理、缺陷偵測和框選,而我們在影像前處理中設計了一個棒鋼邊緣校直的演算法,具有回復棒鋼影像原貌的效果且能幫助提升後續缺陷偵測的正確率;特徵擷取模組則是對框選好的缺陷進行特徵群的記錄,並對特徵群進行分析以選出對分類有用的特徵;缺陷分類模組會將偵測得到的缺陷歸類到其所屬類別,並會根據所要分類的缺陷類別,選取不同的特徵群來提升分類的準確率。我們研究了目前現有的分類技術,決定採用倒傳遞類神經網路(Back-Propagation Neural Network, BPN)以及相關向量機(Relevance Vector Machine, RVM)作為整個系統的分類機制。
在偵測方面,我們請專家用人眼判定缺陷偵測的完整與否,來衡量缺陷偵測的正確性,結果顯示我們設計的偵測演算法能快速並準確地偵測出缺陷所在位置;而在分類方面,我們設計一個階層式的分類架構,並比較BPN與RVM的分類效能,結果顯示RVM具有較為優異的穩定性和準確率。整體來說,我們的系統能快速地偵測缺陷所在,並準確地辨識其為何種缺陷。
This thesis proposes a computer vision system for automatic steel surface inspection. The system analyzes the sequentially acquired steel bar images to detect different kinds of defects on the steel surface. It then classifies the detected defects into the correct categories. The proposed computer vision system has three modules: (1) defect segmentation, (2) feature extraction, and (3) defect classification. The defect segmentation module aims at finding the position of defects and then outlining defects. This module consists of image preprocessing, defect detection, and defects outlining. In the image preprocessing part, we design a boundary adjustment algorithm for the steel bar images. And this algorithm restores the bar image to the original shape to facilitate the subsequent processing and to improve the accuracy of defect detection. The feature extraction module extracts a set of feature values which were defined by experts to be useful for classification from the outlined defects. Finally, the defect classification module classifies the detected defects to the correct categories. In order to achieve the best classification results, this module selects different feature sets at different classification stage of the classification tree according to the defect image characteristics of the corresponding stage. We also utilize two existing classification neural networks, back propagation network (BPN) and relevance vector machine (RVM), as the classifiers of whole system.
In detection, some expert detection results were used to judge the correctness of the proposed defect detection. In defect classification, we propose to use a hierarchical architecture and compare the performance of the two neural network classifiers. The results show that the RVM has better stability and accuracy than BPN and the proposed algorithm could detect defects rapidly and precisely.
[1]Maurizio Valle, Daniela Baratta and Daniele D. Caviglia, ‘Application of Hierarchical Neural Networks to Pattern Recognition for Quality Control Analysis in Steel-Industry Plants,’ International Workshop on Neural Networks for Identification, Control, Robotics, and Signal/Image Processing, pp. 246-252, 1996.
[2]Kyoung Min Kim, Byung Jin Lee, Kyoung Lyou and Gwi Tae Park, ‘Design of a binary decision tree using the genetic algorithm and K-means algorithm for recognition of the defect patterns of cold mill strip,’ IEEE International Fuzzy Systems Conference Proceedings, vol. 2, pp. 1081-1085, 1999.
[3]P. Caleb and M. Steuer, ‘Classification of surface defects on Hot Rolled Steel Using Adaptive Learning Methods,’ Fourth International Conference on Knowledge-Based Intelligent Engineering Systems and Allied Technologies, vol. 1, pp. 103-108, 2000.
[4]Hongbin Jia, Yi Lu Murphey, Jianjun Shi and Tzyy-Shuh Chang, ‘An Intelligent Real-time Vision System for Surface Defect Detection,’ 17th International Conference on Pattern Recognition, vol. 3, pp. 239-242, 2004.
[5]Keesug Choi, Kyungmo Koo and Jin S. Lee, ‘Development of Defect Classification Algorithm for POSCO Rolling Strip Surface Inspection System,’ SICE-ICASE International Joint Conference, pp. 2499-2502, 2006.
[6]Ge-Wen Kang and Hong-Bing Liu, ‘Surface Defects Inspection of Cold Rolled Strips Based on Neural Network,’ Proceedings of the Fourth International Conference on Machine Learning and Cybernetics, pp. 5034-5037, 2005.
[7]Takao Sugimoto and Tadao Kawaguchi, ‘Development of a Surface Defect Inspection System Using Radiant Light from Steel Products in a Hot Rolling Line,’ IEEE Trans. on Instrumentation and Measurement, vol. 47, no. 2, pp. 409-416, 1998.
[8]Se Ho Choi, Jong Pil Yun, Boyeul Seo, YoungSu Park and Sang Woo Kim, ‘Real-Time Defects Detection Algorithm for High-Speed Steel Bar in Coil,’ Enformatika Trans. on Engineering, Computing and Technology, vol. 19, pp. 66-70, 2007.
[9]R.C. Gonzalez and R.E. Woods, Digital Image Processing Second Edition, Prentice Hall, 2002.
[10]M. Bajcsy, M. Mintz, and E. Liebman, ‘A common framework for edge detection and region growing,’ Univ. Pennsylvania GRASP Lab, Technical report 61, 1986
[11]張斐章、張麗秋,類神經網路,臺灣東華書局股份有限公司,pp. 118-143,2005.
[12]Freeman, J. A. and D. M. Skapura, Neural Networks. Algorithms, Applications, and Programming Techniques, Addison-Wesley, Reading, Michigan, 1992.
[13]Michael E. Tipping, ‘Sparse Bayesian Learning and the Relevance Vector Machine,’ Journal of Machine Learning Research, pp. 211-244, 2001.
[14]Michael E. Tipping, ‘The Relevance Vector Machine,’ Advances in Neural Information Processing Systems, pp. 652-658, 2000.
[15]Michael E. Tipping, Sparse Bayesian Learning and the Relevance Vector Machine. Microsoft Research Tech. Rep., Cambridge, U.K. [Online]. Available: http://www.research.microsoft.com/MLP/RVM
[16]B.E. Boser, I.M. Guyou and V. Vapnik, ‘A Training Algorithm for Optimal Margin Classifiers,’ Proc. 5th Ann. Workshop on Computational Learning Theory, pp. 144-152, 1992.
[17]N. Cristianini, J. Shawa-Taylor, ‘An Introduction to Support Vector Machines,’ Cambridge University Press 2000, 2000.
[18]B. Scholkopf, C. J. C. Burges and A. J. Smola, editors. Advances in Kernel Methods: Support Vector Learning, MIT Press, 1999.
[19]C. J. C. Burges, ‘Simplified support vector decision rules,’ In L. Saitta, editor, Proceedings of the Thirteenth International Conference on Machine Learning, pp. 71-77, Bari, Italy, 1996. Morgan Kaufmann.
[20]C. J. C. Burges and B. Scholkopf, ‘Improving the accuracy and speed of support vector machines,’ In M. C. Mozer, M. I. Jordan, and T. Petsche, editors, Advances in Neural Information Processing Systems 9, pp. 375-381. MIT Press, 1997.
[21]J. O. Berger, Statistical decision theory and Bayesian analysis, Springer, second edition, 1985.
[22]D. J. C. MacKay, ‘Bayesian interpolation,’ Neural Computation, 4(3):415-447, 1992.
[23]D. J. C. MacKay, ‘The evidence framework applied to classification networks,’ Neural Computation, 4(5):720-736, 1992.
[24]I. T. Nabney, ‘Efficient training of RBF networks for classification,’ In Proceedings of the Ninth International Conference on Artificial Neural Networks (ICANN99), pp. 210-215, IEE, 1999.
[25]K. V. Mardia, J. T. Kent and J. M. Bibby, Multivariate Analysis (Probability and Mathematical Statistics), Academic Press, 1979.