No description available.

Wai Yin Yee = 遺傳演算法的應用連續影像之人體動作捕捉 / 韋燕儀. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2003. / Includes bibliographical references (leaves 113-115). / Text in English; abstracts in English and Chinese. / Wai Yin Yee = Yi zhuan yan suan fa de ying yong lian xu ying xiang zhi ren ti dong zuo bu zhuo / Wei Yanyi. / Abstract --- p.ii / 摘要 --- p.iv / Acknowledgement --- p.vi / Content --- p.vii / List of Figures --- p.x / List of Tables --- p.xviii / Chapter Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Human Motion Capture --- p.1 / Chapter 1.1.1 --- Optical Motion Capture --- p.3 / Chapter 1.1.2 --- Monocular Motion Capture --- p.4 / Chapter 1.2 --- Proposed Human Motion Capture System --- p.6 / Chapter 1.3 --- Organization --- p.8 / Chapter Chapter 2 --- Introduction of Genetic Algorithms --- p.10 / Chapter 2.1 --- Traditional Search Methods & Genetic Algorithms --- p.11 / Chapter 2.2 --- Mechanism of Genetic Algorithms --- p.14 / Chapter 2.3 --- A Simple Genetic Algorithm --- p.16 / Chapter 2.3.1 --- Initialization --- p.16 / Chapter 2.3.2 --- Evaluation --- p.17 / Chapter 2.3.3 --- Selection --- p.18 / Chapter 2.3.4 --- Genetic Operation --- p.19 / Chapter 2.3.5 --- Termination --- p.23 / Chapter 2.4 --- Convergence Proof for GA --- p.24 / Chapter 2.5 --- Proposed Modified Genetic Algorithm --- p.26 / Chapter 2.6 --- Effectiveness of the Proposed Modified GA on Function Optimization --- p.28 / Chapter 2.6.1 --- Function 1 - Unimodal function --- p.28 / Chapter 2.6.2 --- Function 2 - Sine function --- p.35 / Chapter 2.6.3 --- Function 3 - Foxhole function --- p.39 / Chapter 2.6.4 --- Function 4 - Discrete function --- p.41 / Chapter Chapter 3 --- Pre-processing I - Articulated Stick Model --- p.44 / Chapter 3.1 --- Background Knowledge of Human Skeleton --- p.44 / Chapter 3.2 --- Simplified Humanoid Articulated Stick Model --- p.44 / Chapter Chapter 4 --- Pre-Processing II - Reference Lengths & 2-D Frame Scale --- p.48 / Chapter 4.1 --- Optimization Approach --- p.54 / Chapter 4.1.1 --- Parameters Range --- p.62 / Chapter 4.1.2 --- GA Formulation --- p.63 / Chapter 4.2 --- Triangulation approach --- p.63 / Chapter 4.3 --- Experiments & Discussion --- p.66 / Chapter 4.3.1 --- Experiment One: Synthetic sequences --- p.67 / Chapter 4.3.2 --- Experiment Two: Real image sequences --- p.71 / Chapter Chapter 5 --- Pre-Processing III - Possible Depths --- p.76 / Chapter Chapter 6 --- Resolving Depth Ambiguity by GA --- p.83 / Chapter 6.1 --- Smoothness Assumption --- p.83 / Chapter 6.2 --- Kinematic Constraint --- p.85 / Chapter 6.3 --- GA Formulation --- p.85 / Chapter 6.4 --- Proposed Constrained GA --- p.86 / Chapter 6.5 --- Implementation and Experiments --- p.87 / Chapter 6.5.1 --- Experiment One: Synthetic sequences --- p.88 / Chapter 6.5.2 --- Experiment Two: Real image sequences --- p.105 / Chapter Chapter 7 --- Conclusion --- p.111 / Bibliography --- p.113 / Appendix A Description of Rotating Angles --- p.116

No description available.

One of the difficulties of human figure tracking is that humans move their bodies in complex, non-linear ways. An effective computational model of human motion could therefore be of great benefit in figure tracking. We are interested in the use of a class of dynamic models called switching linear dynamic systems for figure tracking. This thesis makes two contributions. First, we present an empirical analysis of some of the technical issues involved with applying linear dynamic systems to figure tracking. The lack of high-level theory in this area makes this type of empirical study valuable and necessary. We show that sensitivity of these models to perturbations in input is a central issue in their application to figure tracking. We also compare different types of LDS models and identification algorithms. Second, we describe 2-DAFT, a flexible software framework we have created for figure tracking. 2-DAFT encapsulates data and code involved in different parts of the tracking problem in a number of modules. This architecture leads to flexibility and makes it easy to implement new tracking algorithms.