隨著時代的變遷，自動化控制不僅取代了人們的工作，代替人工生產許多物品，現在機器視覺檢測的進步，更可以取代人力進行各樣產品的檢測，此舉不僅可以提高生產的速度，亦提高了產品的穩定度。
本論文的研究目的是利用機器視覺自動去檢測鋰電池的極片高度是否符合標準，因為若電池極片的高度有過高或是過低的現象發生，代表其極片位置已經偏移甚至超出隔離膜，因而導致斷路的現象發生，影響到產品以及使用者的安全，在本論文當中，利用兩支工業用相機，自電池側邊分別對鋰電池的正極（鋁極）以及負極（銅極）進行極片影像的擷取，接著使用中值濾波方式進行雜訊的減除、再經過影像二值化、影像邊緣強化以及影像增長等處理後，擷取極片影像面積、極片影像寬度、極片最高位置以及最低位置四樣比例特徵，輸入倒傳遞類神經網路進行訓練以及影像分類，本論文設計之倒傳遞類神經網路有4個輸入、4個隱藏層以及3個輸出，使用100組正極與100組負極的極片影像進行訓練及分類的測試。分類的輸出有正常、內彎曲以及外彎曲3種項目，分類的目的是為了將因為有彎曲現象的極片挑選出來， 因為這樣的極片影像是無法正確量測其極片高度的。最後將分類為正常的極片影像對其作量化，並與標準值進行比對，若高於標準值則為不良品，反之則為良品。

Throughout the changing times, automation not only replaced people's work but also created many products. Machine vision inspection can replace the human to detect all kinds of products.It not only can increase production speed, but also improve the quality of the products.

The purpose of this thesis is the use of automated machine vision to detect the flatness of the lithium battery pole piece that meets the criteria. If the battery pole piece height is too high or too low, it might cause battery short and affect product and user safety.

This thesis uses two cameras to capture positive (aluminum pole) and negative (copper electrode) pole piece of lithium batteries. It uses median filter to decrease noise and then do the binary, image edge enhancement and image growth. The key parameters include four features that are image area ratio, width ratio, highest position ratios and lowest position ratios of the pole pieces. Then, the calculated data are input to the back-propagation neural network to do the training process and image classification. This neural network has four inputs, four hidden units and three outputs. A hundred positive and a hundred negative pole piece images are used for experiments in this study. The classification outputs include normal, inside and outside bending. The purpose of classification is to locate bending pole piece, because these kinds of images are unable to measure properly.

Finally, the examined product which is good or not good can be determined by comparing with the standard value.

[1] 中國產業研究報告網, www.chinairr.org. (2013)
[2] L. Li-zong, Z. Luo-shan, W. Jian-ding, and Q. Zhi-qin, "Study of PCB Automatic Optical Inspection System Based on Mathematical Morphology," in Computer Technology and Development, 2009. ICCTD '09. International Conference on, 2009, pp. 405-408.
[3] O. Hecht and G. Dishon, "Automatic optical inspection (AOI) of thin film multichip modules," in Electronic Manufacturing Technology Symposium, 1990, IEMT Conference., 8th IEEE/CHMT International, 1990, pp. 130-136.
[4] L. Kunyan and D. Tian, "Research and design of inspection of LR6 battery negative surface scratches online defects based on computer vision," in Communications, Circuits and Systems (ICCCAS), 2013 International Conference on, 2013, pp. 120-123.
[5] 鍾國亮, 影像處理與電腦視覺第三版, 東華書局, 2005.
[6] 井上誠喜，八木伸行，林正樹，中須英輔，三谷公二，奧井誠人, C語言數位影像處理, 全華圖書, 2005.
[7] N. Otsu, "A Threshold Selection Method from Gray-Level Histograms," in Systems, Man and Cybernetics, IEEE Transactions on, 1979, pp. 62 - 66.
[8] G. Wenshuo, Z. Xiaoguang, Y. Lei, and L. Huizhong, "An improved Sobel edge detection," in Computer Science and Information Technology (ICCSIT), 2010 3rd IEEE International Conference on, 2010, pp. 67-71.
[9] A. Tripathi, D. Chourshiya, and Y. Kumar, "Study of image processing in robot visioning," in Control, Automation, Communication and Energy Conservation, 2009. INCACEC 2009. 2009 International Conference on, 2009, pp. 1-4.
[10] J. Canny, "A Computational Approach to Edge Detection," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. PAMI-8, pp. 679-698, 1986.
[11] M. Alasdair, "Introduction to Digital Image Process with MATLAB, 1st ed. Course Technology," 2004.
[12] C. Virginio and P. Alfredo, "Neural Recognition in a Pyramidal Structure," in IEEE Trans. On Neural Networks, Vol.13, no.2,pp.472-480, Mar.2002.
[13] 王進德, 類神經網路與模糊控制理論入門與應用」, 全華圖書, 2011.
[14] 蔡孟凱, 雷穎傑, 黃昭雄, 陳錦輝, and 陳正凱, C++Builder6完全攻略, 金禾資訊, 2007.