本篇論文提出兩套針對不同類型螺帽進行表面缺陷檢測的系統，分別對兩種不同類型的螺帽做檢測。輪緣螺帽分檢系統在整個檢測流程中分為兩大階段：缺陷候選區域偵測和分類。缺陷候選區域偵測的目的為得知缺陷在影像中的可能位置，偵測的方法包含了影像前處理、缺陷偵測和標記。在影像前處理的部分，我們用濾波器將雜訊濾除。接著利用邊緣偵測、邊界搜尋與曲率計算等方式，尋找影像中的缺陷候選區域並予以標記。在分類方面，我們對已標記的缺陷候選區域進行特徵擷取，且針對不同類型的影像分別訓練不同參數的倒傳遞類神經網路分類器。由實驗結果得知，倒傳遞類神經網路分類正確率可達九成，顯示系統可以自動且快速的對輪緣螺帽偵測與分類。
在扣件表面缺陷檢測系統中，首先我們會先定義影像檢測範圍，以節省後續檢測時間。我們用邊緣偵測、連通區域編碼與邊界搜尋等方式尋找螺紋。對於螺紋斷裂(不連續)的缺陷，我們比對編碼一致性來確認缺陷是否存在；對於螺紋不平滑的缺陷，我們計算每一條螺紋本身的曲率與螺紋的平均寬度來尋找缺陷。結合兩種類型的檢測結果，平均檢測正確率可達98%以上，顯示我們的演算法對於自動缺陷偵測相當敏感且深具效率。

This thesis proposes two computer vision defect detection systems for two kinds of nuts respectively. In the flange nut sorting system, the proposed method consists of two phases: detection of the defect candidate and defect classification. In the detection phase, image preprocessing and defect detection are used to find the position of the defect candidate. Image filtering is used to remove the noise in image preprocessing. Then tracing, and curvature method are applied to find the defect candidate. In the classification phase, the back-propagation neural network (BPN) is trained with different parameters respectively. According to the experimental results, the average accuracy rate of classification reaches 90%, and the proposed system is found capable of detecting and classifying defects rapidly and precisely.
In fastener surface defect detection system, the detection regions are empirically defined at the first step to improve the time performance. Then edge detection, connected component labeling and tracing are used to find threads. For broken threads, the label consistency test is used to detect the defect. And for rugged threads, curvature calculation and thickness evaluation are applied to find the defect position. Based on the experiments of the two types of defect detections, the average accuracy rate is more than 98%. It shows that the proposed systems can perform automatic inspection efficiently and accurately.

[1] H. Y. T. Ngan, et al., "Motif-based defect detection for patterned fabric," Pattern Recognition, vol. 41, pp. 1878-1894, 2008.
[2] H. Y. T. Ngan and G. K. H. Pang, "Regularity analysis for patterned texture inspection," IEEE Transactions on Automation Science and Engineering, vol. 6, pp. 131-144, 2009.
[3] H. Y. T. Ngan, et al., "Performance evaluation for motif-based patterned texture defect detection," IEEE Transactions on Automation Science and Engineering, vol. 7, pp. 58-72, 2010.
[4] H. G. Bu, et al., "Fabric defect detection based on multiple fractal features and support vector data description," Engineering Applications of Artificial Intelligence, vol. 22, pp. 224-235, 2009.
[5] O. G. Sezer and A. Ercil, "Using perceptual relation of regularity and anisotropy in the texture with independent component model for defect detection," Pattern Recognition, vol. 40, pp. 121-133, 2007.
[6] S. W. Lin, et al., "Irregular shapes classification by back-propagation neural networks," The International Journal of Advanced Manufacturing Technology, vol. 34, pp. 1164-1172, 2007.
[7] K. Choi, et al., "Development of defect classification algorithm for POSCO rolling strip surface inspection system," presented at the SICE-ICASE International Joint Conference, Korea, 2006.
[8] M. Sonka, et al., Image Processing Analysis, and Machine Vision, Thomson Learning, 2007.
[9] S. G. Tyan, "Median Filtering: Deterministic Properties," in Two-Dimensional Digital Signal Prcessing II. vol. 43, ed Berlin: Springer Verlag, 1981, pp. 197-217.
[10] 蔡翰進, 膝關節X光影像之電腦測量與分析系統, 國立成功大學資訊工程所碩士論文, 2007.
[11] A. Rosenfeld and E. Johnston, "Angle detection on digital curves," IEEE Transactions on Computers, vol. 22, pp. 875-878, 1973.
[12] F. Rosenblatt, "The perceptron: a probabilistic model for information storage and organization in the brain," Psychological Review, vol. 65, pp. 386-408, 1958.
[13] D. E. Rumelhart and J. L. McClelland, Parallel Distributed Processing: Explorations in the Microstructure of Cognition vol. 1, MIT Press Cambridge, MA, USA, 1986.
[14] 張斐章, et al., 類神經網路理論與實務, 東華書局, 2003.
[15] J. A. Freeman and D. M. Skapura, Neural Networks: Algorithms, Applications, and Programming Techniques, Addison-Wesley, 1991.
[16] V. Kecman, Learning and Soft Computing: Support Vector Machines, Neural Networks, and Fuzzy Logic Models, MIT Press Cambridge, MA, USA, 2001.
[17] D. Chester, "Why two hidden layers are better than one," in IEEE International Joint Conference on Neural Networks, Wachington, DC., 1990, pp. 265-268.
[18] V. Kurkova, "Kolmogorov's theorem and multilayer neural networks " Neural Networks, vol. 5, pp. 501-506, 1992.
[19] S. Abe, Support Vector Machines for Pattern Classification, Springer-Verlag, London, 2005.
[20] V. Vapnik, The Nature of Statistical Learning Theory, Springer-Verlag, New York, 1995.
[21] V. Vapnik, Statistical Learning Theory, John Wiley & Sons, New York, 1998.
[22] C. Cortes and V. Vapnik, "Support-vector networks " Machine Learning, vol. 20, pp. 273-297, 1995.
[23] R. C. Gonzalez and R. E. Woods, Digital Image Processing, Prentice Hall, 2002.
[24] A. Rosenfeld and J. S. Weszka, "An improved method of angle detection on digital curves," IEEE Transactions on Computers, vol. 24, pp. 940-941, 1975.