数据驱动的大规模森林场景真实感动画
Data-Driven Realistic Animation of Large-Scale Forest
-
摘要: 提出一种基于数据驱动的大规模森林场景真实感动画的实现方案.该方案预先对单株样本树在风力作用下的运动作高精度的动力学计算,得到高分辨率、精确的树骨架形变序列;然后采用分簇主元分析方法对形变序列数据进行压缩与简化,构建单株树各枝干运动的细节层次;最后基于多株样本树多个细节层次的运动数据合成大规模森林的群体运动.实验结果表明:该方案实现了力学形变数据的高效重用,在森林动画合成效率和真实感上具有一定的优势,能满足对规模和画质要求较高的应用,如影视动画与媒体广告等的需求.Abstract: A data-driven scheme to animate the dynamic swaying of large-scale forest in the wind is proposed.We first calculate the accurate deformations of exemplar trees under different wind conditions by dynamical finite element method to obtain the high resolution displacement sequences of each skeletal segment of those trees at pre-processing stage.Then we conduct clustered principal component analysis to compress and simplify the displacement data of the skeleton segments of each tree.Subsequently,multiple levels of motion details(LOMD)of those trees are constructed with well-adjusted parameters.The dynamic swaying of large-scale forest is finally animated by synthesizing the motion of each tree with view-dependent LOMD.Our experiments demonstrate that the proposed scheme gains both efficiency and realism and thus is favorable for producing high-quality video and media advertisements.
下载: