進階搜尋


 
系統識別號 U0026-0812200914313686
論文名稱(中文) 應用支撐向量迴歸於估測產品造形意象
論文名稱(英文) Support Vector Regression Applied to Estimating Product Form Images
校院名稱 成功大學
系所名稱(中) 工業設計學系碩博士班
系所名稱(英) Department of Industrial Design
學年度 96
學期 2
出版年 97
研究生(中文) 方信恩
研究生(英文) Hsin-En Fang
電子信箱 p3694407@mail.ncku.edu.tw
學號 p3694407
學位類別 碩士
語文別 英文
論文頁數 107頁
口試委員 口試委員-蕭世文
口試委員-孫永年
指導教授-謝孟達
中文關鍵字 產品意象  造形特徵  產品造形  核心函數  支撐向量迴歸  模型建構程序  混合核心函數  感性工學 
英文關鍵字 kernel function  form feature  modeling procedure  product image  product form  Kansei engineering  mixed kernel  support vector regression 
學科別分類
中文摘要 在本篇研究中,我們使用了一個在近年來被廣為發展的新興學習機器理論“支撐向量迴歸”(Support vector regression, SVR) ,來增進電腦系統進行產品造形意象值的估測效能。另外,一種被稱為“核心函數”(Kernel function)的技術也被使用來強化支撐向量迴歸模型處理非線性資料的能力。在本研究中,兩組由虛擬三維模型經過特定程序處理所取得,具有不同資料結構的樣本被用來作為實驗案例以驗證支撐向量迴歸理論的有效性。在第一個研究案例中,32組模擬實際的吹風機產品外形並使用不同的模型建構程序與特徵所建構的虛擬三維模型,被用以作為訓練與測試SVR迴歸模型的實驗樣本。為了統一樣本的資料結構,本研究提出一個前置處理程序,將不同的三維模型皆先轉換成單一的UV舉昇曲面(lofting surface),再於曲面上陣列散佈相同數量的資料點來代表每一個三維模型。之後,我們使用“流線型”這個形容詞語彙來評估這些不同的樣本。而在第二個研究案例中,51組利用特徵導向的造形漸變技巧所產生的三維模型,被用以當作具有系統化結構的資料樣本來訓練與測試SVR迴歸模型。為了建構最佳的迴歸模型,本研究使用“二階段格網搜尋法” (Two-step cross-validation) 來找尋最佳的參數組合。最後,使用了不同的核心函數所建構的SVR迴歸模型表現了良好的估測能力,並顯示出支撐向量迴歸理論的優勢。而兩組具備不同資料結構的樣本也被用來解釋迴歸模型的特性。
英文摘要 In this study, a recently evolved learning machine theory called support vector regression (SVR) is introduced. It is intended to improve the estimating performance of product form image values using a computer system and kernel functions, which are applied to the SVR to increase its ability to resolve nonlinear problems. Two types of samples, which have different data structures derived from virtual 3-D models, are taken as training data for two case studies intended to evaluate the performance of the SVR models. In case study Ι, 32 virtual 3-D models from actual hairdryers are constructed as samples by using different modeling procedures and features. To use these models as input data for training, a process is provided to unify the data structure so as to reconstruct each model as a single UV lofting surface and array points along the surface to provide the data representing an entire 3-D model. Then, a linguistic ‘streamline’ is applied to estimate the power of the SVR. In case study Π, 51 3-D models are prepared as data with systematic structures for testing the SVR using feature-based morphing. To construct the optimal regression model, a technique called two-step cross-validation is used to select the optimal parameter combinations of the SVR model. Finally, the excellent training and predicting power of the SVR model using different kernel functions show the advantages of the SVR approach. The characteristics of the SVR are explained using two types of samples with different data structures.
論文目次 THESIS CERTIFICATION....................................................................................................I
摘要(ABSTRACT IN CHINESE)........................................................................................ II
ABSTRACT ........................................................................................................................ III
ACKNOWLEDGEMENTS ................................................................................................IV
LIST OF TABLES...............................................................................................................XI
CHAPTER 1 INTRODUCTION........................................................................................ 1
CHAPTER 2 RELATED RESEARCH LITERATURES................................................... 5
2.1 Survey of Kansei engineering.................................................................................... 5
2.1.1 Expert systems based on Kansei engineering ................................................... 5
2.2 Product form design using virtual 3-D models .......................................................... 6
2.2.1 Feature-based morphing.................................................................................... 6
2.2.2 Morphing based on figures................................................................................ 7
2.3 Techniques for curved surface construction............................................................... 8
2.3.1 Surface based on NURBS ................................................................................. 8
2.3.2 UV-loft surface .................................................................................................. 9
2.4 Estimation of the product form image ..................................................................... 10
2.4.1 Automatic product form synthesis based on modeling features...................... 10
2.4.2 Product form image analysis based on points ................................................. 11
2.4.3 Technique for product form image estimation ................................................ 12
2.5 Support vector regression ........................................................................................ 14
CHAPTER 3 STRUCTURE OF THEORY...................................................................... 18
3.1 Support vector machine theory ................................................................................ 18
3.2 Kernel functions....................................................................................................... 21
3.3 Support vector regression ........................................................................................ 22
3.4 Local and global kernels .......................................................................................... 30
3.5 Mixtures of kernels .................................................................................................. 33
3.6 Selecting optimal parameter combinations.............................................................. 35
CHAPTER 4 PROCEDURE AND STRUCTURE OF RESEARCH .............................. 38
4.1 Implementation of SVR by computing.................................................................... 38
4.2 Preparing samples for case studies .......................................................................... 39
4.3 Constructing samples for case study Ι ..................................................................... 40
4.3.1 Digitalizing product forms as similar virtual 3-D models............................... 40
4.3.2 Portraiture of product forms by arraying points on UV lofting surface.......... 43
4.3.3 Experiment to obtain product form image ...................................................... 47
4.4 Constructing samples for case study Π.................................................................... 48
4.4.1 Creating 3-D models with feature-based morphing ........................................ 48
4.4.2 Experiment to obtaining product form image ................................................. 50
CHAPTER 5 CASE STUDY Ι ......................................................................................... 52
5.1 Samples for training................................................................................................. 52
5.2 Constructing the SVR training models .................................................................... 52
5.2.1 Training the SVR model using a polynomial kernel....................................... 53
5.2.2 Training the SVR model using an RBF kernel................................................ 65
5.2.3 Training the SVR model using mixed kernels ................................................ 78
5.3 Conclusion: Comparing the three SVR models ....................................................... 82
CHAPTER 6 CASE STUDY Π ...................................................................................... 84
6.1 Experimental samples .............................................................................................. 84
6.2 Constructing the SVR training models .................................................................... 85
6.2.1 Training the SVR model using a polynomial kernel....................................... 85
6.2.2 Training the SVR model using an RBF kernel................................................ 87
6.2.3 Training the SVR model using mixed kernels ................................................ 88
6.3 Estimation of constructed SVR models ................................................................... 90
CHAPTER7 CONCLUSION AND SUGGESTION........................................................ 92
7.1 Conclusion ............................................................................................................... 92
7.2 Suggestion ............................................................................................................... 92
REFERENCES.................................................................................................................... 94
APPENDIX 1 .................................................................................................................... 100
參考文獻 [1] Belaziz, M., Bouras, A., Brun, J.M., “Morphological analysis for product design,” Computer-Aided Design, Vol.32, No.5, pp.377-388, 2000.
[2] Chapelle, Olivier, Haffner, Patrick, and Vapnik, Vladimir N., “Support vector machines for histogram-based image classification,” IEEE Transactions on Neural Networks, Vol.10, No.5, pp.1055-1064, 1999.
[3] Chen, Kuoh-Siang, “Form Generation And Style Association,” Department of Industrial Design, Illinois Institute of Technology, PhD Thesis, 1995.
[4] Desai, Kiran, Badhe, Yogesh, Tambe, Sanjeev S., Kulkarni, Bhaskar D., “Soft-sensor development for fed-batch bioreactors using support vector regression,” Biochemical Engineering Journal, Vol.27, No.33, pp.225-239, 2006.
[5] Fletcher, R., “Practical methods of optimization,” John Wiley and Sons, Inc., 2nd Edition, 1987.
[6] Gustavo, Camps-Valls, Luis, Gómez-Chova, Jordi, Muñoz-Marí, Joan, Vila-Francés, Julia, Amorós-López, Javier, Calpe-Maravilla, “Retrieval of oceanic chlorophyll concentration with relevance vector machines,” Remote Sensing of Environment, Vol.105, Issue 1, pp. 23-33, 2006.
[7] Lin, H.T., Lin, C.J., “A study on sigmoid kernels for SVM and the training of non-PSD kernels by SMO-type methods,” Technical report, Department of Computer Science &Information Engineering, National Taiwan University, 2003.
[8] Han, S. H., Hong, S. W., “A systematic approach for coupling user satisfaction with product design,” Ergonomics, Vol.46, No.13-14, pp.1441-1461, 2003.
[9] Hsiao, S. W., Liu, M. C., “A morphing method for shape generation and image prediction in product design,” Design Studies, Vol.23, Issue 6,pp.533-556, 2002.
[10] Hsiao, Shih-Wen, Tsai, Hung-Cheng, “Applying a hybrid approach based on fuzzy neural network and genetic algorithm to product form design,” International Journal of Industrial Ergonomics, Vol.35, pp.411-428, 2005.
[11] John Shawe-Taylor, Nello Cristianini, “Kernel methods for pattern analysis,” Cambridge university press, 2004.
[12] Karras, D. A. and Mertzios, B. G., ”Time series modeling of endocardial border motion in ultrasonic images comparing support vector machimes multilayer perceptrons and linear estimation technique,” Measurement, Vol.36, Issue 3-4, pp.331-345, 2004.
[13] Kishore Nara and Li Liao,“ Using extended phylogenetic profiles and support vector machines for protein family classification,” SNPD, Bioinformatics, pp.152-157, 2004.
[14] Lai, J.Y., Ueng, W.D., “Reconstruction of surfaces of revolution from measured points,” Computers in Industry, Vol.47, Issue 2, pp.147-161, 2000.
[15] Liu, Jing-Xu, Li, Jin, Tan, Yue-Jin, “An empirical assessment on the robustness of support vector regression with different kernels,” proceedings of the Fourth International Conference on Machine Learning and Cybernetics, Vol.7, pp. 4289-4294, 2005.
[16] Nagamachi, M., “Kansei Engineering: a new ergonomic consumer-oriented technology for product development,” International Journal of Industrial Ergonomics, Vol.15, pp.13-24, 1995.
[17] Oliveira, Adriano L.I., “Estimation of software project effort with support vector regression,” Neurocomputing, Vol.69, Issue 5, pp.17471753, 2006.
[18] P. T. Lin, “Support vector regression: systematic design and performance analysis”, Unpublished Doctoral Dissertation, Department of Electronic Engineering, National Taiwan University of Science and Technology, PhD Thesis, 2001.
[19] Park, J., Han, S. H., “A fuzzy rule-based approach to modeling affective user satisfaction towards office chair design,” International Journal of Industrial Ergonomics, Vol.34, Issue 1, pp. 31-47, 2004.
[20] Schőlkopf, B., Burges, C. J. C. & Smola, A. J., “Introduction to support vector learning advances in kernel methods-support vector learning,” Cambridge, MA , 1999.
[21] Schőlkopf, B., Smola, A. J., “A tutorial on support vector regression,” Statistics and Computing, Vol.14, No.3, pp.199-222, 2004.
[22] Schőlkopf, B., Smola, A. J., “Learning with kernels,” MIT Press, 2002.
[23] Shieh, M.D., Yang, C.C., “Multiclass SVM-RFE for product form feature selection,” Expert Systems with Applications: An International Journal, Vol.35, Issue 1-2, pp. 531-541, 2008.
[24] Shigeo Abe, “Support vector machines for pattern classification,” Springer, 2005.
[25] Shimizu and Jindo, “A fuzzy logic analysis method for evaluating human sensitivities,” International Journal of Industrial Ergonomics, Vol.15, pp. 39-47, 1995.
[26] Smith, G.F., Jordan, E.M, “Improved SVM Regression using Mixtures of Kernels,” IEEEE, Vol.3, pp. 2785-2790, 2002.
[27] Tovey, Michal, “Styling and design: intuition and analysis in industrial design,” Design Studies, Vol.3, Issue 1, pp. 5-31, 1997.
[28] Ueng, W.D., Lai, J.Y., and Doong, J.L., “Sweep-surface reconstruction from three-dimensional measured data,” Computer-Aided Design, Vol.30, No.10, pp.791-805, 1998.
[29] Vapnik V., “Statistical Learning Theory”, John Wiley and Sons, New York, 1998.
[30] Vapnik V., “The Nature of Statistical Learning Theory”, Springer, New York, 1995.
[31] Yang, Shan-Sheng, Lu, Wen-Cong, Chen, Nian-Yi, Hu, Qian-Nan, “Support vector regression based QSPR for the prediction of some physicochemical properties of alkyl benzenes,” Journal of Molecular Structure, Vol.719, Issue 1-3, pp.119-127, 2005.
[32] Zheng, Sheng, Liu, Jian, Tian, Jin-Wen, “An efficient star acquisition method based on SVM with mixtures of kernels,” Pattern Recognition Letters, Vol.26, Issue 2, pp.147-165, 2005.
[33] 李怡君,“交叉路口交通量預測模式之研究”,台灣大學土木工程研究所碩士論文,2004。
[34] 周俊宏,“運用類神經網路與支撐向量機於個人信用卡授信決策之研究”,台灣科技大學資訊管理研究所碩士論文,2006。
[35] 林大仁, “運用支撐向量迴歸與倒傳遞類神經網路建構產品造形設計之預測模式”,成功大學工業設計研究所碩士論文,2007。
[36] 林佳駿,“應用逆向工程於產品快速設計之直接塑模”,成功大學工業設計研究所碩士論文,2002。
[37] 林宜賢,“應用平滑支撐向量迴歸與類神經網路於共同基金績效之預測”,國立臺灣科技大學資訊管理研究所碩士論文,2007。
[38] 林政謙,“應用平滑支撐向量機與類神經網路預測台灣上市電子指數漲跌之研究”,東吳大學經濟所碩士論文,2006。
[39] 馬居正, “應用類神經網路於汽車造形特徵輔助設計之研究”,成功大學工業設計研究所碩士論文,2006。
[40] 張鴿,陳書開, “基於SVM的手寫體阿拉伯數字識別”, 軍民兩用技術與產品,第05期,2005。
[41] 曹鈞斐,“運用平滑支撐向量機於期貨市場擇時策略之研究”,國立台灣科技大學資訊管理研究所碩士論文,2007。
[42] 莊峻超, “A Study on Appling Reverse Engineering to Morphing Shape Design”, 成功大學工業設計研究所碩士論文,2002。
[43] 陳奕仲, “應用遺傳演算法於電腦輔助數位相機造型設計”,成功大學工業設計研究所碩士論文,2004。
[44] 陳寬裕,“結合遺傳演算法與支援向量迴歸於台灣股票加權指數之預測”,計量管理期刊,第3卷第1期,1-18頁,2006。
[45] 黃俊明,“逆向工程曲面重建與特徵變更之研究”,台灣科技大學機械工程研究所博士論文,1998。
[46] 黃科森,“臉部表情之合成與變形”,清華大學資訊工程學系碩士論文,2001。
[47] 黃敏菁,“支撐向量機在財務時間序列預測之應用”,輔仁大學金融研究所碩士論文,2005。
[48] 黃銘智,“逆向工程-結合擷取設計特徵重建幾何模型之資料點前處理方法”,大同大學機械工程研究所博士論文,2000。
[49] 詹佳尉, “應用凸輪運動方程式於汽車造形設計”,成功大學工業設計研究所碩士論文,2002。
[50] 劉孟昌,“形態漸變之產品造型衍生與整合模式研究”,成功大學工業設計研究所碩士論文,1999。
[51] 劉瓊山, “舉昇曲面之逆向工程研究”,中央大學機械工程研究所碩士論文,1995。
[52] 鄭忠杰, “特徵參數化與網際網路於產品造形衍生之研究”,成功大學工業設計研究所碩士論文,2003。
[53] 魏志驊, “應用人工智慧之機車造型自動化設計系統”,成功大學工業設計研究所碩士論文,2002。
論文全文使用權限
  • 同意授權校內瀏覽/列印電子全文服務,於2010-08-22起公開。
  • 同意授權校外瀏覽/列印電子全文服務,於2010-08-22起公開。


  • 如您有疑問,請聯絡圖書館
    聯絡電話:(06)2757575#65773
    聯絡E-mail:etds@email.ncku.edu.tw