基于子结构的植物功能结构随机模型

CORC > 自动化研究所 > 中国科学院自动化研究所 > 毕业生 > 博士学位论文

题名	基于子结构的植物功能结构随机模型
作者	康孟珍
学位类别	工学博士
答辩日期	2003-06-01
授予单位	中国科学院研究生院
授予地点	中国科学院自动化研究所
导师	胡包钢
关键词	植物模型功能结构模型随机过程子结构快速算法概率泰勒级数递归 Plant model Functional structural model stochastic process substructure fast algorithm mean and variance probability Taylor
其他题名	Functional and Structural Stochastic Plant MOdeling based on Substructures
学位专业	模式识别与智能系统
中文摘要	植物生长建模是植物学、数学、计算机科学等多种学科相结合的课题。最近几年中，植物模型研究的方向是将植物的内在的生理过程和空间形态相结合的功能结构模型。GreenLab模型是近年来一直在发展着的一个植物结构功能模型，主要包括形态发生模型和生理模型两部分；计算的结果为单株植物在各个周期产生的生物量、器官的大小、方向和空间位置等。以往的GreenLab模型是个确定性模型，不能反映植株个体之间的差异。本文的主要研究内容是在GreenLab模型中引入模拟芽的生长活动的随机过程，并预测和模拟随机植物结构中器官个数和产量的统计特征。 GreenLab模型中的形态发生模型为一个双尺度自动机，其中用"宏状态"、"微状态"及相互之间的跳转关系模拟植物的生长单元、生长单元中的叶元以及它们的生长顺序。在确定性生长的情况下，由于植物的分层组织特性，可以通过子结构方法可以递归地算出由形态发生模型得到的器官（节间、花、叶、果）的个数。计算所得的器官的个数将与GreenLab模型的生理模型的参数相结合，例如器官的相对汇强、扩张函数、异速生长系数等，用于计算每个生长周期总的生物量产量及分配，各种器官的重量、大小等等。根据得到的器官大小以及各种角度的定义可以构造出植物的三维形态。在GreenLab确定性模型框架基础之上，本文在形态发生模型中引入与芽的生长活动相联系的各种生长概率，即植物顶芽的存活概率、生长概率、叶元的出现概率和侧芽的分枝概率等。这样，每个周期的器官个数以及产生的生物量的变化都变成随机过程。为计算随机生长下的生物量的均值与方差，本文首先利用复合过程的求解方法以及子结构方法逐步地推导出分枝结构中器官个数的均值、方差和协方差，然后将结果用于各个生长周期随机生物量的均值、方差与协方差的递推计算。为验证上述理论计算结果，观察各种概率对植物拓扑结构的影响，本文的另一方面工作是随机植物结构的模拟。由于结构复杂的树结构一般包含大量的器官，逐节地模拟多个随机树的样本会占用大量计算机资源。因此本文提出用随机子结构的方法构造树结构，即对于每种属性的子结构，构造一个随机样本集，每一个子结构样本的生长轴用逐节的蒙特卡罗方法模拟，轴上的子结构从对应的已生成的样本集中随机选取，不再重新模拟。样本集的大小会影响到模拟的精度与效率。由于对多棵植物只需构造一次样本集，因此可以提高复杂植物以及植物群落的构造
英文摘要	Plant growth modeling is a pluridisciplinary subject including botany, mathematics and computer science. In the recent years, the main direction of research on plant model is Functional- Structural Model(FSM) that integrate the intrinsic physiological process of plant with architectural model. GreenLab model is a FSM that was developing in recent years, including a morphogenesis model and a physiological model; the result of model includes biomass production, organ weight, direction and position in each cycle of a single plant. Previously GreenLab model is deterministic and can't reveal the difference between plants. The content of this thesis is to introduce stochastic process that simulate activities of buds, and then compute mean and variance of organ number and biomass production in branching structure which were compared to results from simulation. The morphogenesis model of GreenLab is a Dual-scale automaton, in which 'macrostate' and 'microstate' and their transition rules are used to simulate growth units, metamers inside each growth unit and their growth sequence individually. Thanks to the hierarchical structure of plant, in case of deterministic growth, the number of organs from morphogenesis model can be calculated recursively with substructure method. Combined with parameter of physiological model like relative sink strength, expansion law and allometric parameters, the computed organ number can be used to calculate total biomass production and allocation in each cycle, as well as weights and sizes of organs. With these data and some predefined angles the 3D shape of plant can be constructed. Based on framework of GreenLab deterministic model, in this thesis, some growth probabilities related with bud activities were introduced into the morphogenesis model. Thus organ number and biomass production in each cycle become stochastic process. In order to compute mean and variance of biomass production in each cycle, firstly the mean, variance and covariance of new organ number in each cycle are deduced with method of compound process and substructure. These results are applied further more to deduce mean, variance and covariance of biomass production in different cycle. To verify result of theoretical computation and to observe effect of probabilities on plant structure, another work of this thesis is the simulation of stochastic plant structure. Since a complex tree structure usually includes a lot of organs, to simulate numerous stochastic tree samples can have high costs in computer resources. Hereby in this thesis, method of stochastic substructure is proposed; that is, to each kind of substructure, a set of stochastic samples are created; for each sample structure, only the bearing axis is built with Monte-Carlo method, while the substructures on the axis are selected randomly from already built set. The size of set will influence preciseness and efficiency of simulation. Since the se
语种	中文
其他标识符	837
内容类型	学位论文
源URL	[http://ir.ia.ac.cn/handle/173211/5771]
专题	毕业生_博士学位论文
推荐引用方式 GB/T 7714	康孟珍. 基于子结构的植物功能结构随机模型[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2003.