基于视觉运动捕获数据的三维模型驱动方法

Method of driving a 3D model based on vision-based motion capture data

基于视觉运动捕获数据的三维模型驱动涉及到将三维位置数据转化为模型运动数据,和三维模型的动作驱动。视觉运动捕获的三维运动数据,由于缺乏骨架各关节点间的旋转信息,不能直接用于三维模型的驱动。本文提出了依据各关节点的相关性,求解骨架关节点的旋转信息,并计算出各关节点的欧拉角,将原始据转换为模型运动数据。然后,将转换后的运动数据应用于三维模型,通过改进后的线性混合皮肤变形算法(LBS),实现了基于三维运动参数的三维模型驱动。文中实验结果验证了算法的有效性。

The raw data of motion capture based on vision can not drive a 3D model directly, due to lac- king of rotation information between skeleton joints. A new method is presented for solving the problem of acquiring the rotation information of each joint depends on the relativity of skeleton joints. Then, to calcu- late the Euler angle of eaeh joint. The motion data including Euler angle is employed on driving a 3D model through algorithm improved linear blend skinning （LBS）. Finally, a 3D human model is driven au- tomatically. At the end of this paper, several experiments are given to demonstrate that the proposed method is efficient.