四元数样条插值的人体运动数据重构

Human motion data reconstruction using quaternion spline interpolation

为了得到平滑的人体动画,提出一种基于四元数的样条插值算法,利用提取的关键帧实现人体运动序列的有效重构。为减少重构误差、加快收敛速度,将已知关键帧集合作为初始条件,通过迭代算法求出样条曲线的控制点集合。利用样条曲线控制点计算贝塞尔曲线控制点,构造贝塞尔样条曲线段,将各段贝塞尔样条曲线段组合,构造一条基于四元数的样条曲线。根据德卡斯特里奥(de Casteljau)算法插值重构人体运动。实验结果表明,该算法在保证执行效率的同时,可得到光滑的插值结果,实现满足视觉要求的人体运动重构。

To obtain smooth keyframes-based human motion animation,an interpolation algorithm of quaternion spline is presented,which constructs a smooth interpolation curve on the 4D sphere by using ideas of spline.An iterative method is utilized to compute control points of quaternion spline,a collection of keyframes as initial approximation is selected in order to minimize reconstructed error and maximize the convergence.A segment Bezier curve is computed through four control points of quaternion spline,as a result that a quaternion spline is a composite curve consisting of several Bezier curve segments.The de Casteljau algorithm is used as a method of evaluating quaternion spline to interpolate.Experimental results show the presented algorithm can obtain real-time smooth interpolation,and the reconstructed motion is closely to the ideal.