植被是户外场景的重要组成部分，植被绘制效果良好可以增强三维虚拟场景的逼真度。由于植被结构复杂、形态多变，因此对森林景观进行完全细节的描述是难以实施的。如何描绘出大规模的森林景观，使其既具有一定逼真度又能满足实时交互是多年来国内外研究者关注的焦点。目前提高大规模森林场景实时绘制效率的方法主要包括两方面：一方面是提高硬件水平，即采用处理速度更快的计算机图形硬件来加速渲染；另一方面是对场景本身的绘制算法进行改进和完善。目前计算机的计算能力往往无法快速真实地模拟大规模森林，随着计算机图形技术的发展，需要更快更真实地构建大规模场景，单纯依赖于硬件加速的绘制技术是非常有限的，因此需要从图形绘制算法着手，提高实时绘制的效率。本研究采用L系统构建植被单体的三维结构模型，根据植被模型特点，分别对枝干和树叶两个组成部分进行几何简化，大幅度减少模型所需的几何要素数量。在实时渲染过程中，依据细节层次（LOD）技术对植被进行不同程度的简化并显示，从而实现大规模森林场景的实时可视化。在枝干模型简化中，本文在多分辨率字符串算法的基础上，探索了分枝方式对LOD模型的可视化效果的影响，改进了合轴分枝枝干系统的简化方法。在冠层模型简化中，本文提出了一种新的分层LOD模型提取方法，该方法可以高效地提取出视相关的多复杂度LOD模型。多复杂度LOD模型是指同一个LOD模型分为细节区和粗糙区，两个区域的复杂度不同。这种模型可以更进一步地减少几何绘制量。经实验验证该方法具有良好的显示效果和运算效率。在森林场景可视化中，本文提出了一种新的动态LOD调用标准：定义衰减因子描述树木分布情况，并用该因子将初始树木的复杂度衰减到显示的需要，从而实时得到场景中植被所需要的复杂度。该算法全面考虑了植被密度和视距对植被所需细节程度的衰减影响，为动态LOD技术运用于森林场景提供了连续的复杂度标准。

The tree plant family occupies a great area of the world’s land surface, and it’s essential to most nature scenes. However, rendering detailed trees is a huge challenge in terms of rendering speed because of its geometric complexity. So the visualization technology of forestall plant, is the core technology in the computer graphic field, and it has a very high academic significance and the application value.The models are formed by large number of polygons. The large amount of geometry necessary to simulate natural environments causes many problems at rendering time. Not even powerful graphic accelerators are able to generate high enough frame rates for the interactive rendering of the scene. To reduce the amount of geometry rendered we need software-based strategies, like multiresolution modeling.This paper uses random-parametric L-systems integrated with field survey data to construct 3D tree models. According to the structure of trees, we implement the geometric simplification respectively for the branches and leaves to dramatically reduce the polygon count. The continuous level-of-detail (LOD) method is implemented to rendering the plants with multiresolution models.In the branch model simplification, this paper simplified the tree based on in the multi-resolution algorithm. The branching way of the tree exerts the influences on the visualization effects, so we improved the simplification method for the trees with sympodial branching.In the simplification of canopy models, this paper presents a hierarchical extraction method of the level-of-detail (LOD) model. This approach can effectively build variable canopy’s LOD models, and guarantee each leaf in the LOD models minimizes the error equation. By the variable LOD model we refer to the model adapt multiresolution of the different parts of the foliage, thus allowing the higher resolution in the zone of interest than in the coarsest zone. This paper also presents a new set of criteria for running the LOD vegetation models: A decay factor is introduced to describe the distribution of trees and is used to decide the rendering complexity of vegetation in the scene in real time. The algorithm considers the influences, which are exerted by the density of vegetation and the distance from the viewpoint to the vegetation, on the degradation to the details of trees. In this way, our approach provides continuous complexity of the LOD model for dynamic LOD visualization of trees in large forest landscapes. The experiments show that the presented method here has better visual quality and high rendering efficiency.