臺灣博碩士論文加值系統

傳統植物辨識的研究都是以整張圖像進行特徵提取，且這些特徵多半是基於植物結構上的形態為主。利用整張圖取特徵，容易造成特徵可能來自於背景的雜訊，同時會耗費過多的計算時間。另外形態特徵容易受到生物變異性和環境因素的影響，因此較為不穩定。
本研究透過GrabCut將背景濾除，接著擷取葉子所在的區域，再分別利用深度信念網路(Deep Belief Net，DBN)對圖像自動學習特徵，以及人工提取詞袋特徵(Bag of Words，BoW)結合支持向量機(Support Vector Machine，SVM)來進行植物的辨識。根據特徵提取實驗的比較，我們先擷取葉子所在的區域再取特徵，確實能夠縮短特徵提取的計算時間。另外辨識率的部份，以人工提取Bag of Words特徵最高的辨識率可達95.31%，而利用DBN學習特徵所得到的辨識率則為75%。

Traditional plant recognition researches use whole-plant images as source of feature extraction, and the extracted features are mostly morphologically based. Using whole-plant images means more unrelated background noises and computation time; besides, morphologically based features may easily be influenced by biological variations and environmental factors, which means the extracted features are not stable.
In this research, we filter out the background by using GrabCut, then extract bounding rectangle of the plant. Two machine learning methods are then used to construct the plant recognition model. The first method directly uses the above extracted images as training input for Deep Belief Net to automatically learn the features; as for the second method, Bag-of-Word features are further extracted and applied to Support Vector Machine. According to the experimental results, we can see the extraction of specific plant area can really improve the computation time of feature extraction. As for the recognition experiments, when SVM along with Bag-of-Word features is used, the best accuracy we can get is 95.31% while the other approach using DBN can give an accuracy of 75%.

謝誌 I
摘要 II
Abstract III
目次 IV
圖目次 VI
表目次 VIII
第一章 緒論 1
1-1研究動機 1
1-2研究目的 2
1-3論文結構 3
第二章 相關研究 4
2-1 GrabCut 4
2-1-1顏色模型 5
2-1-2迭代能量最小化 6
2-1-2交互式操作與不完全的標註 7
2-2 二值化 8
2-3 輪廓表示 11
2-4 尺度不變特徵轉換 12
2-4-1尺度空間極值點的偵測 12
2-4-2關鍵點定位 14
2-4-3關鍵點之方向定位 14
2-4-4關鍵點描述 15
2-5 K-means 分群演算法 16
2-6 Bag-of-Words模型 17
2-7 支持向量機 19
2-7-1 線性可分 19
2-7-2 線性不可分 20
2-7-3 非線性支持向量機 22
2-8 深度學習 24
2-8-1 限制性波茲曼機 24
2-8-2 深度信念網路 28
2-9 植物辨識文獻探討 29
第三章 研究方法 30
3-1 影像前處理 31
3-1-1 RGB轉換灰階 31
3-1-2 GrabCut濾除背景 32
3-1-3 擷取感興趣的區域 35
3-2 特徵提取 37
3-2-1 Bag of SIFT特徵 37
3-3 DBN學習特徵 39
3-4 支持向量機進行分類 41
第四章 實驗結果與分析 42
4-1 實驗環境與資料庫 42
4-2 Bag of Word結合SVM實驗結果 46
4-3 DBN實驗結果 49
4-4 分析與比較 51
第五章 結論與展望 53
參考文獻 54

[1] Botanic Gardens Conservation International,“Plant Species Numbers”, https://www.bgci.org/ourwork/1521/
[2] Current Results weather and science facts, “Total Number of Species Estimated in the World”,
http://www.currentresults.com/Environment-Facts/Plants-Animals/estimate-of-worlds-total-number-of-species.php
[3] Leaf structure, http://imgarcade.com/1/leaf-structure-diagram-to-label/
[4] C. Rother, V. Kolmogorov, and A. Blake, “GrabCut – interactive foreground extraction using iterated graph cut,” ACM Trans. On Graphics –Proc. Of ACM SIGGRAPH, vol. 23, no. 3, pp. 309-314, August 2004.
[5] Y. Y. Boykov, M. P. Jolly, “Interactive Graph Cuts for Optimal Boundary and Region Segmentation of Objects in N-D Images,” International Conference on Computer Vision, In ICCV, vol. 1, pp. 105-112, July 2001.
[6] N. Otsu, “A Threshold Selection Method from Gray-Level Histograms,” IEEE Trans. On System, Man and Cybernetics, vol. SMC-9, pp.62-66, January 1979.
[7] S. Suzuki, K. be, “Topological structural analysis of digitized binary images by border following,” Computer Vsion, Graphics, and Image Processing, vol. 30, Issue 1, pp. 32-46, April 1985.
[8] D. G. Lowe, “Distinctive Image Features from Scale-invariant Keypoints,” International Journal of Computer Vision, vol. 60, no. 2, pp.91-110, November 2004.
[9] J. B. MacQueen, “Some methods for classification and analysis of multivariate observation,” Proc. Fifth Berkeley Symp. On Math. Statist. And Prob., vol. 1, pp. 281-297, 1967.
[10] G. Csurka, C. R. Dance, L. Fan, J. Willamowski, C. Bray, “Visual Categorization with Bags of Keypoints,” Workshop on statistical learning in computer vision, ECCV, vol. 1, 2004.
[11] Li Fei-Fei, P. Perona, “A Bayesian Hierarchical Model for Learning Natural Scene Categories,” IEEE CVPR, pp. 524-531, 2005.
[12] V. N. Vapnik, “The nature of statistical learning theory,” Springer, 1995.
[13] C. Cortes, V. Vapnik, “Support-vector Networks,” Machine Learning, vol. 20, pp. 279-297, 1995.
[14] G. E. Hinton, S. Osindero, Yee-Whye Teh, “A fast learning algorithm for deep belief nets,” Neural Computation, vol. 18, Issue 7, pp. 1527-1554, July 2006.
[15] G. E. Hinton, “Training Products of Experts by Minimizing Contrastive Divergence,” Neural Computation, vol. 14, Issue 8, pp. 1771-1880, August 2002.
[16] J. S. Cope, D. Corney, J. Y. Clark, P. Remagnino, and P. Wilkin, “Plant species identification using digital morphometrics: A review,” Expert Systems with Application, vol. 39, No. 8, pp. 7562-7573, June 2012.
[17] S. G. Wu, F. S. Bao, E. Y. Xu, Yu-Xuan Wang, Yi-Fan Chang and Qiao-Liang Xiang, “A Leaf Recognition Algorithm for Plant Classification Using Probabilistic Neural Network”, IEEE International Symposium on Signal Processing and Information Technology, pp. 11-16, December 2007.
[18] A. Kadir, L. E. Nugroho, A. Susanto, and P. I. Santosa, “Foliage plant retrieval using polor fourier transform, color moments and vein features,” Signal and Image Processing, vol. 2, no. 3, September 2011.
[19] Z. Y. Wang, B. Lu, Z. Chi, and D. G. Feng, “Leaf Image Classification with Shape Context and SIFT Descriptors,” International conference on Digital Image Computing：Techniques and Application, pp. 650-654, December 2011.
[20] S. Mouine, I. Yaniaoui, A. Verroust-Blondet, “A shape-based approach for leaf classification using multi-scale triangular representation”, ACM International Conference on Multimedia Retrieval, pp. 127-135, April 2013.
[21] A. H. Kulkarni, H. M. Rai, K. A. Jahagirdar, and P. S. Upparamani, “A leaf recognition technique for plant classification using RBPNN and Zernike moments,” Advanced Research in Computer and Communication Engineering, vol. 2, no. 1 , pp. 984-988, January 2013.
[22] K. Mahdikhanlou and H. Ebrahimnezhad, “Plant leaf classification using centroid distance and axis of least inertia method,” The 22nd Iranian Conference on Electrical Engineering, pp.1690-1694, May 2014.
[23] B. Oluleye, A. Leisa, D. Dean, and J. S. Leng, “A Neuronal Classification System for Plant Leaves Using Genetic Image Segmentation,” Mathematics and Computer Science, vol. 9, no. 3, pp. 261-278, May 2015.
[24] DBN Image-Classification,
https://github.com/lazyprogrammer/image-classification-dbn
[25] C. C. Chang, C. J. Lin, “LIBSVM—A Library for Support Vector Machine,” Available: http://www.csie.ntu.edu.tw/~cjlin/libsvm/index.html
[26] C. W. Hsu, C. C. Chang, C. J. Lin, “A Practical Guide to Support Vector Classification,” April 2010.
[27] M. F. Kazerouni, J. Schlemper, K. D. Kuhnert, “Comparison of Modern Description Methods for the Recognition of 32 Plant species,” Signal and Image Processing, vol. 6, no. 2, pp. 1690-1694, April 2015.
[28] Introduction to Computer Vision ,“Definition of Bag of Word”,
http://vision.stanford.edu/teaching/cs231a_autumn1112/lecture/lecture15_bow_part-based_cs231a_marked.pdf
[29] C. J. C. Burges, “A Tutorial on Support Vector Machine for Pattern Recognition,” Data Mining and Knowledge Discovery, pp.121-167, 1998.