利用力学预计算方法生成植物运动的真实感动画需要对形变数据作有效简化和压缩。针对植物及其运动天然具有层级性的特点，提出一种基于视觉感知层次的动态几何简化算法。依据植物的树型层级结构，将各级“子树”的运动作高低频分解；“低频运动”表征“子树”的主体运动，“高频运动”即为运动细节。按“广度优先”遍历植物层级结构，在某一“深度”（层级）上将所有“子树”的“低频运动”合成即获得原始运动在此层次的逼近。遍历深度递增将构成植物运动的渐进逼近；递减则为植物运动的逐步抽象化，这实质上实现了植物运动的细节层次控制。实验表明该算法可实现植物运动的高效简化和压缩，利于生成大规模植物场景的真实感动画。

Decimation and compression of pre-computed deformation data is crucial to synthesize realistic animation of massive plants. In this paper, a motion simplification algorithm is proposed based on the observation that plants are of tree-like hierarchical structure inherently and move in a hierarchical mode. According to the hierarchy of a plant, the motions of its sub-trees in each depth are decomposed into low-frequency motion and high-frequency motion. The former denotes the dominant transformation of the plant within that sub-tree range and the latter contains residual deformation details. By breadth-first traversal of the hierarchy, the motion composed with low-frequency motions of all the sub-trees at a certain depth is the approximation of original motion at that depth. While traversal depth increased, the approximated motion would approach the original motion progressively and become more realistic, which results in a kind of levels of motion detail . The experimental results show that the presented scheme can simplify and compress the plants motions efficiently, and is convenient for the realistic animation of a large-scale forest.