臺灣博碩士論文加值系統

香蕉黃葉病是香蕉栽培歷史上最具破壞性的疾病，亦有「香蕉癌症」之稱。隨著黃葉病迅速蔓延，造成全世界最具經濟價值之一的香蕉嚴重減產，臺灣蕉農亦深受其害，故需要及時發現及處理，避免其擴散。近幾年，許多研究應用深度學習於影像辨識上面，結果非常優異，故本研究亦以深度學習之卷積神經網路進行香蕉黃葉病偵測，使用五種較知名卷積神經網路架構搭配七種激勵函數組合，訓練並比較三十五種搭配組合之優劣，提供一個可用於辨識香蕉黃葉病的卷積神經網路最佳架構。實驗結果顯示，使用ResNet-34搭配ELUs激勵函數達到99.72%正確率，具有高精準度的黃葉病辨識能力。

Panama Disease, also known as the cancer of banana, is the most devastating disease in the banana cultivation. Due to the pervasion of Panama Disease all around the world, the number of bananas, one of the cash crops in this world, has severely plummeted, and banana farmer in Taiwan also become the victims under the spread of this disease. Therefore, the signs of this disease must be found and dealt with in time to prevent the disease from spreading. In recent years, many researches utilize deep-learning image recognition, and the outcomes of these researches are extraordinary. Thus, this study also employs deep-learning Convolutional Neural Network (CNN) to execute the detection of Panama Disease. By using five well-known structures of CNN, along with the combination of seven Activation Functions, and training as well as comparing the advantage and disadvantage of 35 combinations, the best CNN structure of detecting Panama Disease is discovered. The result shows that the combination of ResNet-34 and ELUs Activation Function can achieve 99.72% accuracy; hence, it has high-precision on recognizing Panama Disease

目錄
摘要.....................................................I
Abstract ................................................II
謝誌.....................................................III
目錄.....................................................IV
圖目錄...................................................VIII
表目錄...................................................XI
第壹章 緒論..............................................1
第一節 研究背景與動機....................................1
第二節 研究流程..........................................4
第三節 研究範圍與限制....................................5
第四節 論文架構..........................................5
第貳章 文獻探討..........................................7
第一節 香蕉介紹..........................................7
第二節 香蕉黃葉病........................................7
第三節 彩色影像..........................................10
一、 像素 ...............................................10
二、 影像尺寸 ...........................................10
三、 位元深度 ...........................................11
四、 色彩三原色 .........................................12
第四節 人工神經網路......................................13
第五節 深度學習..........................................17
一、 卷積層 .............................................18
二、 線性整流層 .........................................19
三、 池化層 .............................................28
四、 全連接層 ...........................................30
五、 損失函數層 .........................................30
第六節 卷積神經網路代表性架構............................31
一、 LeNet ..............................................31
二、 AlexNet.............................................32
三、 VGGNet..............................................33
四、 ResNet..............................................35
第七節 卷積神經網路應用於農業領域........................39
第參章 研究方法..........................................40
第一節 香蕉黃葉病辨識研究流程............................40
第二節 方法概念..........................................42
一、 預處理資料集 .......................................42
二、 卷積神經網路 .......................................43
第三節 香蕉黃葉病辨識步驟................................48
一、 香蕉黃葉病辨識之網路模型訓練步驟....................49
二、 香蕉黃葉病辨識測試步驟 .............................50
第四節 評估指標..........................................51
第肆章 實驗結果與分析....................................53
第一節 實驗環境..........................................53
第二節 香蕉黃葉病資料集..................................54
第三節 模型與參數設定....................................56
第四節 實驗結果..........................................59
一、 以 LeNet-5 為架構對各激勵函數之分析.................59
二、 以 VGG-16 為架構對各激勵函數之分析..................60
三、 以 VGG-19 為架構對各激勵函數之分析..................62
四、 以 ResNet-34 為架構對各激勵函數之分析...............63
五、 以 ResNet-50 為架構對各激勵函數之分析...............64
第伍章 結論與未來展望....................................66
第一節 結論..............................................66
第二節 未來展望..........................................68
參考文獻.................................................69
附錄.....................................................78
附錄一 香蕉黃葉病辨識系統操作圖..........................78
附錄二 各網路模型與激勵函數執行電腦運算次數30次之結果....82
作者簡介.................................................117

[1]A. Pereira and M. Maraschin, “Banana (Musa spp) from peel to pulp: Ethnopharmacology, source of bioactive compounds and its relevance for human health,” Journal of Ethnopharmacology, vol. 160, pp. 149–163, 2015.
[2]M. K. Smith, A. W. Whiley, C. Searle, P. W. Langdonb, B. Schaffer, and K. G. Pegg, “Micropropagated bananas are more susceptible to fusarium wilt than plants grown from conventional material,” Australian Journal of Agricultural Research, vol. 49, no. 7, pp. 1133–1139, 1998.
[3]R. C. Ploetz, “Panama Disease : A Classic and Destructive Disease of Banana,” Plant Health Progress, vol. 10, pp. 1–7, 2008.
[4]R. Ploetz and K. Pegg, “Fusarium wilt of banana and Wallace’s line: Was the disease originally restricted to his Indo-Malayan region?,” Australasian Plant Pathology, vol. 26, no. 4, pp. 239–249, 1997.
[5]S. DeOliveira, M. Teixeira, S. Junior, É. J.Alves, andJ.Raimundo, “Banana breeding program at Embrapa,” Crop Breeding and Applied Biotechnology, vol. 1, no. 4, pp. 399–436, 2001.
[6]R. A. Meldrum, A. M. Daly, L. T. T. Tran-Nguyen, and E. A. B. Aitken, “Are Banana weevil borers a vector in spreading Fusarium oxysporum f. sp. cubense tropical race 4 in banana plantations?,” Australasian Plant Pathology, vol. 42, no. 5, p. 543, 2013.
[7]J. Amara, B. Bouaziz, and A. Algergawy, “A Deep Learning-based Approach for Banana Leaf Diseases Classification,” pp. 79–88, 2017.
[8]K. O’Donnell, H. C. Kistler, E. Cigelnik, and R. C. Ploetz, “Multiple evolutionary origins of the fungus causing Panama disease of banana: concordant evidence from nuclear and mitochondrial gene genealogies.,” Proceedings of the National Academy of Sciences of the United States of America, vol. 95, no. 5, pp. 2044–2049, 1998.
[9]W. H. Ko and S. C. Hwang, “Cavendish Banana Cultivars Resistant to Fusarium Wilt Acquired through Somaclonal Variation in Taiwan,” Plant Disease, vol. 88, no. 6, pp. 580–588, 2004.
[10]L. Pérez-vicente, M. A.Dita, and E. Martínez De La Parte, “Technical Manual: Prevention and diagnostic of Fusarium Wilt(Panama disease) of banana caused by Technical Manual Prevention and diagnostic of Fusarium Wilt (Panama disease) of banana caused by Fusarium oxysporum f . sp . cubense Tropic,” Worshop on Diagnosis of Fusarium Wilt, vol. 4, no. May, pp. 1–74, 2014.
[11]W. S. McCulloch and W. Pitts, “A logical calculus of the ideas immanent in nervous activity,” The Bulletin of Mathematical Biophysics, vol. 5, no. 4, pp. 115–133, 1943.
[12]D. O. Hebb, “The Organization of Behavior: A Neuropsychological Theory,” Organization, p. 62, 1949.
[13]F. Rosenblatt, “The perceptron: A probabilistic model for information storage and organization in the brain,” Psychological Review, vol. 65, no. 6, pp. 386–408, 1958.
[14]J. J. Hopfield, “Neural networks and physical systems with emergent collective computational abilities.,” Proceedings of the National Academy of Sciences, vol. 79, no. 8, pp. 2554–2558, 1982.
[15]D. E. Rumelhart, J. L. McClelland, and C. PDP Research Group, Eds., Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1: Foundations. Cambridge, MA, USA: MIT Press, 1986.
[16]S. B. Kotsiantis, “Supervised Machine Learning: A Review of Classification Techniques,” Informatica, vol. 31, pp. 249–268, 2007.
[17]L. D. LeCun Jackel, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, B.LeCun, J.Denker, andD.Henderson, “Handwritten Digit Recognition with a Back-Propagation Network,” Advances in Neural Information Processing Systems, pp. 396–404, 1990.
[18]C. Cortes and V. Vladimir“Support-vector networks.” Machine learning , vol. 20, no. 3, pp. 273-297, 1995.
[19]T. Kohonen, “The self-organizing map,” Neurocomputing, vol. 21, no. 1–3, pp. 1–6, 1998.
[20]G. A. Carpenter, S. Grossberg, and D. B. Rosen, “Fuzzy ART: Fast stable learning and categorization of analog patterns by an adaptive resonance system,” Neural Networks, vol. 4, no. 6, pp. 759-771, 1991.
[21]J. Acquarelli, T.van Laarhoven, J. Gerretzen, T. N. Tran, L. M. C. Buydens, and E. Marchiori, “Convolutional neural networks for vibrational spectroscopic data analysis,” Analytica Chimica Acta, vol. 954, pp. 22–31, 2017.
[22]S. Bai, “Growing random forest on deep convolutional neural networks for scene categorization,” Expert Systems with Applications, vol. 71, pp. 279–287, 2017.
[23]C. Spampinato, S. Palazzo, D. Giordano, M. Aldinucci, and R. Leonardi, “Deep learning for automated skeletal bone age assessment in X-ray images,” Medical Image Analysis, vol. 36, pp. 41–51, 2017.
[24]X. W. Gao, R. Hui, and Z. Tian, “Classification of CT brain images based on deep learning networks,” Computer Methods and Programs in Biomedicine, vol. 138, pp. 49–56, 2017.
[25]R. M. Menchón-Lara, J. L. Sancho-Gómez, and A. Bueno-Crespo, “Early-stage atherosclerosis detection using deep learning over carotid ultrasound images,” Applied Soft Computing Journal, vol. 49, pp. 616–628, 2016.
[26]S. Afaq Ali Shah, M. Bennamoun, and F. Boussaid, “Iterative deep learning for image set based face and object recognition,” Neurocomputing, vol. 174, pp. 866–874, 2016.
[27]D. Silver, A. Huang, C. J. Maddison, A. Guez, L. Sifre, G. Van Den Driessche, and S. Dieleman,“Mastering the game of Go with deep neural networks and tree search,” Nature, vol. 529, no. 7587, pp. 484–489, 2016.
[28]G. Cybenko, “Approximation by superpositions of a sigmoidal function,” Mathematics of control, signals and systems, vol. 2, no. 4, pp. 303–314, 1989.
[29]J. Han and C. Moraga, “The influence of the sigmoid function parameters on the speed of backpropagation learning,” in International Workshop on Artificial Neural Networks, 1995, pp. 195–201.
[30]X. Glorot, A. Bordes, and Y. Bengio, “Deep sparse rectifier neural networks,” in Proceedings of the fourteenth international conference on artificial intelligence and statistics, 2011, pp. 315–323.
[31]V. Nair and G. E. Hinton, “Rectified linear units improve restricted boltzmann machines,” in Proceedings of the 27th international conference on machine learning (ICML-10), 2010, pp. 807–814.
[32]B. Xu, N. Wang, T. Chen, and M. Li, “Empirical evaluation of rectified activations in convolutional network,” arXiv preprint arXiv:1505.00853, 2015.
[33]A. L. Maas, A. Y. Hannun, and A. Y. Ng, “Rectifier nonlinearities improve neural network acoustic models,” in Proc. icml, 2013, vol. 30, no. 1, pp. 3.
[34]K. Konda, R. Memisevic, and D. Krueger, “Zero-bias autoencoders and the benefits of co-adapting features,” arXiv preprint arXiv:1402.3337, 2014.
[35]D. A. Clevert, T. Unterthiner, and S. Hochreiter, “Fast and accurate deep network learning by exponential linear units (elus),” arXiv preprint arXiv:1511.07289, 2015.
[36]G. Klambauer, T. Unterthiner, A. Mayr, and S. Hochreiter, “Self-normalizing neural networks,” in Advances in Neural Information Processing Systems, 2017, pp. 971–980.
[37]Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-Based Learning Applied to Document Recognition,” PR OC. OF THE IEEE, NOVEMBER, pp. 1–46, 1998.
[38]A. Krizhevsky, I. Sutskever, and G. E.Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” Advances In Neural Information Processing Systems, pp. 1–9, 2012.
[39]K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition,” pp. 1–14, 2014.
[40]K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778.
[41]M. Dyrmann, H. Karstoft, and H. S. Midtiby, “Plant species classification using deep convolutional neural network,” Biosystems Engineering, vol. 151, no. 2005, pp. 72–80, 2016.
[42]J. L. Tang, D. Wang, Z. G. Zhang, L. J. He, J. Xin, and Y. Xu, “Weed identification based on K-means feature learning combined with convolutional neural network,” Computers and Electronics in Agriculture, vol. 135, pp. 63–70, 2017.
[43]E. Jahani Heravi, H. Habibi Aghdam, and D. Puig, “An optimized convolutional neural network with bottleneck and spatial pyramid pooling layers for classification of foods,” Pattern Recognition Letters, vol. 0, pp. 1–9, 2017.
[44]S.Wu, S. Zhong, and Y. Liu, “Deep residual learning for image steganalysis,” Multimedia Tools and Applications, pp. 1–17, 2017.
[45]C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, and A. Rabinovich, “Going Deeper with Convolutions,” arXiv:1409.4842, 2014.
[46]Y. Lu, S. Yi, N. Zeng, Y. Liu, and Y. Zhang, “Identification of rice diseases using deep convolutional neural networks,” Neurocomputing, vol. 267, pp. 378–384, 2017.