Physically-Based Tree Animation and Leaf Deformation Using CUDA in Real-Time

摘要

This paper presents a novel physically-based parallel approach to animate tree motion in realtime and the leaf deformation is accelerated on a CUDA-based platform. Since physically-based tree animation can hardly achieve realtime performance due to the complicated geometry and expensive calculation. Therefore in this paper, three main measures are taken to overcome this problem. Firstly, we briefly introduce a method of physically-based tree motion called hierarchical matrix structure model driven by the external forces such as the wind; then we analyze the model on a parallel platform in detail; finally, all the tree data structures will be redefined as arrays which are suitable for parallel implementation on GPU. In addition, leaf deformation with a double layer structure, caused by its internal forces, will also be well mapped from CPU to GPU using a similar parallel mechanism. Experimental results show that many species of trees can animate realistically and naturally in realtime; Meanwhile, leaf deformation can be plausibly simulated and the performance will be improved by up to ten times.