| Generating face models, facial animations and motion simulation for 3D avatars are currently hot research points in computer graphics field because of their wide applications in games, films and visual communication, etc. This thesis focuses on face modeling, facial animation, robot motion simulation, and the construction of these prototype systems.In the area of face modeling and animation, we implement face modeling system based on single frontal photo and expression animation system based on MPEG-4. Firstly, we capture a frontal photo with ordinary cameras. Secondly, we use active shape models (ASM) algorithm to automatically detect the location of 77 feature points, which well mark facial features of human face, including face contour, eyes, nose, mouth, etc. Thirdly we employ radial basis functions (RBF) interpolation method to deform a neutral generic model to the personalized geometry model according to the detected features. Fourthly, we extract the texture from the photo and do texture mapping to generate realistic model. In the generic model we specify feature vertexes in facial definition parameter (FDP), define facial animation parameter units (FAPU) according to MPEG-4 standards, and then we construct the FaceDefTables by calculating motion factors of vertexes which are affected by every facial animation parameter, last we achieve the realistic facial animation according to the FaceDefTables. Experimental results show that our system can generate realistic face model and facial animation quickly and expediently without any user interaction.In the area of robot motion simulation, we make use of the Microsoft Robotics Studio, which is a platform for robotics development, and on the basis of this platform we develop the motion simulation system for robots, the system first defines a Joint-Rod structure for a given robot model and then drives all joints movement to achieve motion simulation and control using user provided driven data and this Joint-Rod structure. |
PDF Full Text Request