近年來，在資訊化越來越發達的時代，人們以經開始從二維邁向三維的時代了，加上最近的三維影像越來越廣泛的使用，例如像是在卡通、電玩、工程和電影等方面。而那些3D models的創建基本上都是利用一種叫作Motion capture的技術去補捉，而那些設備對於一般使用者太過於昂貴且難以取得，為了解決這問題，我們提出一個新的想法。利用兩台攝影機同步去拍攝一段影片，利用細化與追蹤演算法來達到關節點的取得，且把二維座標轉換成三維座標。
在此篇論文裡，我們結合了細化演算法與追蹤演算法，利用此兩個方法去抓取人體的關節點，再利用兩台攝影機所拍攝出的影像差異來算出每個關節點的三維座標，最後在把這些三維座標輸出至VRML裡，讓這些三維座標重建出三維化身。

Recently, 3D models are widely used in many areas such as cartoons, games, movie industry and engineering, etc. Presently, the most 3D Human kinematic motion models are constructed by optical devices such as 3D scanner, and Motion Capture. Generally, these devices are expensive and difficult to acquire, however. To avoid this problem, we propose a novel approach which based on the multi-cameras tracking to construct the 3D human avatars, since the motion capture devices are too expensive and difficult to acquire.
An adaptive approach to skeleton detection is presented. First, we propose a novel method to combine thinning and tracking algorithm which can find the skeleton tracking points into three-dimensional avatars. The joints can be found by the thinning algorithm. And we also added tracking method to find the joints when the objects are disappeared. Finally, we reconstruct the 3D avatars by using multiple cameras to get the X, Y, and Z-axis coordinate.

目錄
第一章 緒論.....................................1
1.1	研究動機與目的..........................1
1.2	論文組織介紹............................5
第二章 理論基礎與相關研究.......................7
2.1　彩色模組...................................7
2.2　相關影像處理技術..........................14
2.2.1　中位數過濾(Median Filter)：.............18
2.2.2　侵蝕運算(Erosion)：.....................20
2.2.3　擴張運算(Dilation)：....................21
2.2.4　Closing運算：...........................22
2.2.5　Opening運算：...........................23
2.3　國內外相關研究............................24
第三章 研究方法與進行步驟......................26
3.1　細化演算法(Thinning Algorithm)............28
3.1.1　Zhang-Sun演算法.........................31
3.1.2　Rule-Based Thinning演算法...............33
3.2　找尋關節點................................36
3.3　膚色偵測..................................39
3.4　追縱物件..................................41
3.5　利用多攝影機重建三維化身..................48
第四章 系統實作................................53
4.1　系統介面..................................53
4.2　開發平台..................................58
4.3　系統簡介..................................63
4.3.1　程式系統架構............................63
4.3.2　編輯元件介紹............................66
第五章 結論及未來展望..........................81
5.1　結論......................................81
5.2　未來展望..................................83
參考文獻.......................................84
附錄　英文論文.................................86

圖目錄
圖一、Motion Capture所須穿的感應衣..............3
圖二、Motion capture所使用的特殊攝影機..........4
圖三、RGB色彩空間模組...........................8
圖四、CMY顏色空間的表示圖.......................9
圖五、HSI顏色空間色彩分佈示意圖................12
圖六、CCD擷取物件流程圖........................14
圖七、物件擷取範例圖...........................17
圖八、八鄰居圖像...............................18
圖九、中位數過濾的結果.........................19
圖十、Structuring element......................20
圖十一、侵蝕運算的結果.........................20
圖十二、擴張運算的結果.........................21
圖十三、Closing運算的結果......................22
圖十四、Open運算結果...........................23
圖十五、此研究方法的流程圖.....................27
圖十六、典型的細化問題.........................29
圖十七、細化所出現的問題.......................30
圖十八、ZS演算法圖例...........................32
圖十九、Rule-Based Thinning演算法的20 Rules....33
圖二十、Rule-Based Thinning演算法..............34
圖二十一、細化結果比較圖.......................35
圖二十二、端點與分支點圖例.....................37
圖二十三、找尋關節點圖例.......................38
圖二十四、膚色偵測圖例.........................40
圖二十五、物間追蹤演算法流程圖.................42
圖二十六、物件特徵點的圖例。...................44
圖二十七、利用預測位置找尋下一個物件...........46
圖二十八、物件追蹤演算法.......................47
圖二十九、兩台攝影機擺放圖.....................49
圖三十、利用兩台攝影機同步拍攝的影像圖例.......49
圖三十一、攝影機擺放位置圖.....................51
圖三十二、產生VRML三維化身圖例.................52
圖三十三、系統主畫面...........................54
圖三十四、操縱系統之選單.......................54
圖三十五、顯示正在分析影片中的畫面.............55
圖三十六、使用者設定各項的門檻值...............55
圖三十七、顯示影片的圖像檔名...................56
圖三十八、顯示各項資訊視窗.....................57
圖三十九、Borland C++ Builder 6.0軟體介面......59
圖四十、專案導覽區圖示.........................60
圖四十一、元件選取區圖示.......................61
圖四十二、元件編輯區圖示.......................61
圖四十三、系統設計及程式撰寫區圖示.............62
圖四十四、編輯程式元件結構示意圖...............63
圖四十五、系統主畫面...........................64
圖四十六、MainMenu選取位置圖...................68
圖四十七、MainMenu設計圖.......................68
圖四十八、載入圖像檔案對話框...................70
圖四十九、二質化系統介面.......................71
圖五十、中位物過濾的介面.......................71
圖五十一、細化的程式介面.......................72
圖五十二、找出關節點並標上不同數字.............73
圖五十三、利用膚色偵測出來的圖像...............73
圖五十四、物件追蹤系統畫面.....................74
圖五十五、TImage元件選取位置...................75
圖五十六、呈現左方攝影機之圖像.................76
圖五十七、左方圖像運算後的結果.................76
圖五十八、呈現右方攝影機之圖像.................77
圖五十九、右方圖像運算後的結果.................77
圖六十、OpenDialog選取位置圖...................78
圖六十一、Edit元件選取位置圖...................79
圖六十二、Label元件選取位置圖..................79
圖六十三、Memo元件顯示的結果...................80

[1] J W Tukey, “Nonlinear (nonsuperposable) methods for smoothing data”, in Proc. Cong. Rec. EASCON, p. 673. 1974. 

[2] R. M. Haralick, S. R. Stemberg, and X. Zhuang, “Image analysis using mathematical morphology”, in IEEE Trans. Putt. Anal. Machine Intell, vol.PAMl-9, no. 4, pp. 532-550, July 1987

[3] Hwann-Tzong Chen, Horng-Horng Lin, Tyng-Luh Liu,"Multi-object tracking using dynamical graph matching", Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on, Volume: 2 , 8-14 Dec. 2001, Pages:II-210 - II-217 vol.2

[4] Mark C.K. Yang, Jong-Sen Lee, Fellow, IEEE,Cheng-Chang Lien, and Chung-Lin Huang "Hough Transform Modified by Line Connectivity and Line Thickness", IEEE Transactions on Pattern Analysis and Machine Intelligence ,Volume 19 Issue 8 (August 1997, Pages: 905 - 910,Year of Publication: 1997

[5] Y.T. Zhang and C. Suen, “A Fast Parallel Algorithm for Thinning Digital Patterns”, Comm. ACM, vol. 27, no.3, pp. 236-239, 1984.

[6] M J Jones, J M Rehg, ”Statistical color models with application to skin detection”, International Journal of Computer Vision. vol. 1, 1999

[7] Morimoto, T.; Kiriyama, O.; Harada, Y.; Adachi, H.; Koide, T.; Mattausch, H.J, “Object tracking in video pictures based on image segmentation and pattern matching”, ISCAS, vol.4, pp.3215-3218, 2005

[8] Sharples M., Hogg D., Hutchison C., Torrance S. & Young D, “Computers and Thought: An Introduction to Cognitive Science and Artificial Intelligence”, MIT Press - Bradford Books, Chapter 9, 1989 
[9] Ahmed, M.; Ward, R, ”A rotation invariant rule-based thinning algorithm for character recognition”, IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 24, no.12, pp.1672-1678, 2002

[10] Jincheng Wang, Yan Ma, “A face detection method based on color information and eyes characteristics”, Proceedings of the 6th World Congress on Intelligent Control and Automation, 2006

[11] Jinfeng Yang; Zhouyu Fu; Tieniu Tan; Weiming Hu, “Adaptive skin detection using multiple cues”, Image Processing ICIP’04 , vol.2, pp.901-904, 2004

[12] Xiaolin Wu, “YIQ vector quantization in a new color palette architecture”, IEEE Trans. Image Processing, vol. 5, issue 2, pp.321-329, 1996

[13] Kalman, R. E. "A New Approach to Linear Filtering and Prediction Problems," Transactions of the ASME - Journal of Basic Engineering Vol. 82: pp. 35-45, 1960