臺灣博碩士論文加值系統

三維頭部姿勢追蹤與估測在電腦視覺系統之應用中扮演重要角色，如人機互
動介面、表情分析、三維人臉模型重建等方面。本論文目的在改良Jing Xiao 所
提出之三維Lucas-Kanade 追蹤演算法。我們提出了調適性樣版(Adaptive Template)
的技術。調適性樣版上的每個像素點是一個一維高斯模型，當每一個畫面追蹤完
畢時，調適性樣版即以獲得的新的樣本點即時更新各像素點之高斯模型，藉此使
得追蹤演算法更能應用於動態的環境條件之改變，例如光線變動之情況。實驗結
果亦顯示，未使用調適性樣版的追蹤演算法，在光影改變的環境條件下追蹤即告
失敗，然而本研究所提出之方法仍可以持續正確地追蹤頭部姿勢。

3D head tracking and pose estimation play important roles in many vision-based
applications, such as man-machine interaction, facial expression analysis, 3D facial
model reconstruction, etc. This research aims to improve the 3D Lucas-Kanade
tracking algorithm proposed by Jing Xiao. We propose an adaptive template technique
for the Lucas-Kanade tracking algorithm. In this technique, the value of each pixel on
the adaptive template is modeled by a 1-D Gaussian distribution. On finishing the
tracking process of each frame, the adaptive template immediately updates the
Gaussian distributions of the pixels based on the pixels on the newly tracked head
image. In this way, the proposed method can better handle the dynamic changes of the
environment conditions, such as illumination changes. The experimental results show
that the conventional tracking algorithm fails to track the head under varying
illumination, while the proposed tracking algorithm still can track the head and
estimate the pose successfully.

Abstract..................................................................................................................3
摘要........................................................................................................................4
目錄........................................................................................................................5
圖目錄....................................................................................................................7
表目錄....................................................................................................................8
第 1 章 導論........................................................................................................9
1.1 研究動機.................................................................................................9
1.2 相關研究.................................................................................................9
1.2.1 分解法..........................................................................................9
1.2.2 對應點式......................................................................................10
1.2.3 粒子濾波器，與粒子濾波器結合梯度下坡法..........................11
1.2.4 Lucas-Kanade 方法......................................................................12
1.2.5 相關研究總整理..........................................................................12
1.3 系統架構.................................................................................................13
1.4 章節架構.................................................................................................13
第 2 章 三維Lucas-Kanade 頭部追蹤..............................................................15
2.1 二維Lucas-Kanade 追蹤法...................................................................15
2.2 移動模型(motion model)........................................................................17
2.2.1 twists 轉換模型............................................................................17
2.2.2 投影模型......................................................................................18
2.2.3 移動模型......................................................................................19
2.3 三維 Lucas-Kanade 方法......................................................................20
2.3.1 演算法架構..................................................................................20
2.3.2 方程式推導..................................................................................21
2.4 樣版權重.................................................................................................22
2.4.1 可見度(visibility).........................................................................22
2.4.2 影像梯度(image gradient)強度...................................................23
2.4.3 誤差影像(error image) ................................................................23
第 3 章 可調適性模型用於三維LK 追蹤法....................................................25
3.1 可調適性模型.........................................................................................25
3.2 可調適性模型用於三維LK 之推導.....................................................27
3.3 追蹤演算法.............................................................................................29
第 4 章 實驗結果................................................................................................31
4.1 實驗環境.................................................................................................31
4.2 多種樣版之實驗.....................................................................................31
4.3 改變樣版權重之實驗.............................................................................36
4.4 光線改變影像序列追蹤之實驗.............................................................40
4.5 追蹤失敗之討論.....................................................................................46
4.6 實驗討論.................................................................................................47
結論與未來展望....................................................................................................49
參考文獻................................................................................................................51

[1] 黃啟倫。駕駛人視線與瞌睡之自動監控。碩論。June, 2003.
[2] Jing-xiang Chai, Jing Xiao and Jessica Hodgins, Vision-based Control of 3D
Facial Animation. Eurographics/SIGGRAPH Symposium on Computer Animation
(2003). p.193-206.
[3] Jing Xiao and Jinxiang Chai and Takeo Kanade, A Closed-Form Solution to
Non-Rigid Shape and Motion Recovery. The 8th European Conference on Computer
Vision (ECCV 2004), May, 2004. 573-587.
[4] Jing Xiao and Takeo Kanade, Non-Rigid Shape and Motion Recovery: Degenerate
Deformations. Proceedings of the IEEE Conference on Computer Vision and Pattern
Recognition, June 2004. 668-675.
[5] Robert Laganière, Sébastien Gilbert, and Gerhard Roth, Robust Object Pose
Estimation From Feature-Based Stereo, IEEE Transactions On Instrumentation And
Measurement, Vol. 55, No. 4, August 2006.
[6] Qiang Ji and Rong Hu, 3D Face Pose estimation and tracking from a monocular
camera, Image and Vision Computing (2002). pp. 499-511.
[7] F. Dornaika and F. Davoine , Head and Facial Animation Tracking using
Appearance-Adaptive Models and Particle Filters. Conference on Computer Vision
and Pattern Recognition Workshop, 2004. 153- 153.
[8] C. Tomasi and T. Kanade, Shape and motion from image streams under
orthography: a factorization method, International Journal of Computer Vision,
9(2):137-154, 1992.
[9] C. Bregler, et al, Recovering Non-Rigid 3D Shape from Image Streams,
Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2000.
690-696. vol.2.
[10] F. Pighin, et al, Synthesizing Realistic Facial Expressions from Photographs,
Conference on Computer Graphics and Interactive Techniques, 1998. 690 - 696 vol.2
[11] Zhengyou Zhang, et al, Robust and Rapid Generation of Animated Faces from
Video Images: A Model-Based Modeling Approach, International Journal of Computer
Vision, 2004. 93-119. vol58.
[12] P. Yao, et al, Using Affine Correspondence to estimate 3-D Facial Pose,
International Conference on Image Processing, 2001. 852-855
[13] Kenji Oka, et al. Head Pose Estimation System Based on Particle Filtering with
Adaptive Diffusion Control, IAPR Conference on Machine Vision Applications, 2005.
[14] Kenji Oka, et al, Real-time Modeling of Face Deformation for 3D Head Pose
Estimation., Proceedings of the IEEE International Workshop on Analysis and
Modeling of Faces and Gestures, 2005. p.586-589
[15] J. Xiao, et al, Robust Full-Motion Recovery of Head by Dynamic Templates and
Re-Registration Techniques, Proceedings of the Fifths IEEE International Conference
of Automatic Face and Gesture Recognition, 2002. 156-162.
[16]B.D. Lucas. and T. Kanade, An Iterative Image Registration Technique with an
Application to Stereo Vision, Proceedings of the 7th International Joint Conference on
Artificial Intelligence, 1981. 674-679.
[17] Simon Baker and Iain Matthews, Lucas-Kanade 20 Years On: A Unifying
Framework, International Journal of Computer Vision, 2004. 221-255. Vol 56.
[18] Christoph Bregler and Jitendra Malik, Tracking People with Twists and
Exponential Maps. IEEE Computer Society Conference on. Computer Vision and
Pattern Recognition, 1998. 8-15.
[19] Richard M. Murray, Zexiang Li and S.Shankar Sastry, A Mathematical
Introduction to Robotic Manipulation. CRC Press, 1994.
[20] D. Hearn and M.P. Baker, Computer Graphics with OpenGL third edition,
Prentice Hall.
[21] http://www.cs.bu.edu/groups/ivc/data.php
[22] Allan D. Jepson, David J. Fleety and Thomas F. El-Maraghi, Robust Online
Appearance Models for Visual Tracking. IEEE Conference on Computer Vision and
and Pattern Recognition, 2001, Vol. I, pp. 415~ 422.