【作者】 陈立；

【导师】 孙守迁；

【作者基本信息】 浙江大学 ， 计算机应用技术， 2006， 硕士

【摘要】 自从Parke在1972创建了第一个三维人脸表情动画开始,三十年来人脸造型和动画技术已经发展成为两大类,数个小类的综合技术。这些技术分别在各自的领域中得到应用。本论文作为“计算机辅助眼镜设计系统”项目的后续研究,在已有的人脸区域控制模型的基础上进行改进,建立适合表情动画的参数化可变形人脸模型,并通过全局肌肉变形和局部FFD变形,实现了一个人脸三维表情系统。 在已有的人脸区域控制模型的基础上,根据人脸的皮肤、肌肉和骨骼结构,本论文改进后得到的参数化可变形人脸模型由3部分组成:皮肤网格、肌肉和头骨。为了模拟皮肤,在弹性网格上定义了非线性结构弹簧和边界约束弹簧,非线性结构弹簧用来模拟皮肤的弹性,边界约束弹簧用来防止皮肤被拉破;肌肉层根据FACS中对表情变化起决定作用的AU定义相关的肌肉向量,同时将皮肤网格根据不同肌肉的作用范围分块,在肌肉作用范围内定义FFD变形块,用以在产生不真实变形时来对皮肤网格进行精细的调整,而肌肉则分成3类分别模拟:线形肌肉、括约肌和片状肌肉;头骨是可活动的,下颌骨可上下活动以模拟嘴巴的张合,此外头骨层的存在是为了防止在肌肉伸缩的过程中将皮肤网格拉扯到不真实的位置。在对参数化可变形人脸模型进行全局变形时,通过对皮肤网格上的点的受力分析来计算点在某种表情下的位移。皮肤网格上的点受3种力:弹簧内部拉力、肌肉拉力以及头骨的约束力。分别计算3种力的大小,然后使用二阶常微分方程计算某个时刻点的位移,从而得到变形后的表情。之后通过变形率的大小来判断是否对某个区域使用局部FFD变形来修补变形失真或变形不够。 本论文实现的人脸表情方法通过关键人脸表情系统,效果得到了较好的验证。本文将前人的很多工作综合起来,得到参数化可变形人脸模型控制参数少,操纵比较简单,适合快速但对场地和模型精度要求不高的场合应用。更多还原

【Abstract】 Facial modeling and animation has been developed into 2 big categories and several small types integrated together since Parke created the first 3D facial animation in 1972. This thesis is the fellow study of the Computer Aided Glasses Design System. Improved the existent Local Control Model, we present a parameterized deformable facial model to suite the expression animation, and use global muscle deformation and local free-form deformation, implement a 3D facial animation system.We create a parameterized deformable facial model based on the structure of the human face. The model is composed of 3 parts: skin mesh, muscle and skull. To simulating the skin, we define the nonlinear spring and the ERS (Edge Repulsion Spring) on the elastic skin mesh. The ERS is used for preventing the collapse of skin. The muscle is defined by the muscle vector based on the crucial AUs in the FACS, and the skin mesh will divide into areas by the influence of the muscle and define FFD deformation space around these areas to adjust the unrealistic deformation. The muscle is classified into 3 types: linear muscle, sphincter and sheet muscle. The jaw of the skull is rotatable, and the existence of the skull is used for preventing the point to move to the unrealistic place. During the global deforming of the parameterized deformable facial model, we analyze the forces acting on a point and calculate the displacement of this point under a certain expression. The point on the skin mesh is under 3 types of forces: internal force of the spring, muscle force and constrained force of the skull. After computing these forces, we use the second-order ordinary differential equation (ODE) to compute the displacement of the point, and gain an expression. And then, we consider the deformation rate to see if need to use the FFD local deformation.Through the 3D facial animation system, we see a better proof. The thesis integrates many of the methods before. Our model has fewer parameters, and easily to manipulate, suite for quick modeling but not so accuracy.更多还原