中文摘要：

半球成像方法可广泛应用于植被结构参数及冠层光合有效辐射测量等领域。高逼真度虚拟森林环境的半球成像方法模拟及验证可用于定量评估半球成像方法测量精度及地面验证研究。但高逼真度虚拟森林环境下半球成像方法模拟面临场景复杂度高、面片数巨大、模拟效率低等难题。首先，论文以高逼真度虚拟森林环境构建原理建立了不同林分密度、树木分布模式、场景大小及树木种类等特征的虚拟森林环境库；然后，在采用光线跟踪算法开展半球成像方法模拟时，对比分析了4种数据结构分割平面搜索方法，即空间中分法、对象中分法、启发式排序法和启发式BIN分区法，结果表明，启发式排序法性能最佳；最后，以高逼真度虚拟森林环境库，对比分析了传统方法、Kdtree、BVH、Octree4种数据结构方案组织、管理单树几何模型及虚拟森林场景两级数据集效率，并选择场景节点遍历数、单树几何模型节点遍历数、单树几何模型遍历数、面片遍历数及平均计算时间等参数作为衡量数据结构组织方案性能的定量化指标，研究发现Kdtree数据结构方案较其他3种数据结构方案更为优越，因此，其可作为高逼真度虚拟森林环境库的半球成像方法模拟之首选。

英文摘要：

Hemispherical photography method was widely used in the measurement of structural parameters and photosynthetically active radiation of vegatation canopies. The simulation of hemispherical photography method under high-reality virtual forest environment can be used to evaluate the measurement precision of hemispherical photography method quantitatively and used as a useful tool for ground based validation. But there are some dif- ficulties faced for the analogy of hemispherical photography method under high-reality virtual forest environ- ment, such as the highly complex of scene, huge number of triangles, the distribution model of trees, and the low efficiency of simulation etc. In this study, a library of virtual forest scenes with characteristics of different stand density, the distribution mode of tree, size, species etc. was build first based on the principle of the establishment of high-reality virtual forest environment, and then, a comparative analysis between four different splitting plane search methods （such as spatial median algorithm, object median algorithm, heuristic ordering algorithm and heu- ristic BIN zoning algorithm） was carried out based on the ray tracing and hemispherical photography analogy methods. Results show that the heuristic ordering algorithm is the best choice for the analog of hemispherical photography method as compared to the other three splitting plane search methods. At last, the efficiency of the organization of data sets at two scales of isolated tree model and virtual forest scene for four data structures was comparative analyzed based on the library of high reality of virtual forest environment. A list of parameters was used as the indicator to evaluate the performance of the four data structures, such as the number of the ergodic nodes of scene, the number of the ergodic nodes of isolated tree model, the ergodic number of isolated trees, the number of ergodic triangles and the average compute time etc. Results show that the data structure of Kdtree was superior to th