【作者】 李勇；

【导师】 冯仲科； 臧淑英；

【作者基本信息】 北京林业大学 ， 林业装备工程， 2007， 博士

【摘要】 本文以MODIS遥感数据为数据源，运用灰色建模理论，建立林火蔓延的指数模型。本研究的目的在于，利用现代测量技术，建立从遥感数据获取、数据加工、数据处理、林火蔓延模型建立以及林火蔓延模拟的技术体系。通过林火蔓延模型的建立为林火的扑救提供数据支持，为研究林火的蔓延规律探索新的途径。本文主要以黑龙江省黑河市嫩江县嘎拉山2006年的森林大火为研究背景，以高时间分辨率的MODIS遥感图像作为研究对象。通过对现有在轨卫星的比较，选取具有林火通道的MODIS数据作为本文研究的数据源。在论文数据收集过程中，获取了从5月21日至29日该地区的MODIS遥感图像，包括250m、500m、1000m分辨率从可见光到红外波段的遥感图像。本文基于MODIS图像进行研究，试图从动态遥感图像中挖掘出林火蔓延模型。在数据建模过程中，首先对遥感图像进行预处理，对于预处理的遥感图像再进行数据融合处理，对遥感图像进行融合以增加和补充图像的火灾信息。利用ArcGIS软件对融合后的图像进行矢量化处理，将遥感图像信息转化为建立林火蔓延模型的数据信息。在矢量化的过程中，首先是林火蔓延边界重心的确定，将八叉树理念整合到矢量化过程中，使林火的蔓延方向、地理坐标和遥感图像等达到理论与应用习惯的统一。在GM(1，1)的建立中，根据数据的特点，提出了ER算法和LIR算法，保证了林火蔓延模型的建模精度。从建立的灰色模型看，预测相对误差小于1％的占50％，预测相对误差小于5％的占25％，预测相对误差在5～10％的占25％。并选取砍都河、免渡河以及俄罗斯等地的火场数据作为检核，对于林火蔓延模型的预测精度进行检验，满足建模要求。GM(1，N)模型，为林火相关因子的研究提供了新的途径。在计算机模拟与表达方面，首次提出将灰色林火蔓延模型与元胞自动机相结合，使两者优势互补。同时利用ArcEngine开发了林火预测系统，该系统可完成遥感图像的载入、林火边界跟踪、建模数据自动提取、自动预测以及林火蔓延边界和遥感图像的比较。本文的主要创新点包括：(1)首次采用遥感数据，利用灰色理论，进行林火蔓延模型的建立。探索出一条进行林火蔓延与预测的新方法，建立了一条新的林火蔓延模型研建的技术体系。(2)在模型建立的过程中，将稳健估计引入到灰色理论建模，根据建模数据的特点，首次提出了ER算法和LIR算法。(3)在计算机模拟中，首次提出了基于灰色林火蔓延模型的改进的元胞自动机模型的构想。基于MODIS数据建立灰色林火蔓延模型的技术体系研究，是对传统林火蔓延模型建立模式的补充，是在总结前人经验基础上对数据选取和建模理论的有机结合，为林火蔓延模型的建立提供了广阔的渠道。更多还原

【Abstract】 This paper demonstrate how to establish exponent model of forest fire spreading with Gray System modeling theory and MODIS remote sensing data as original data source.The key point of this paper is to establish the technology system of acquiring remote sensing data, polising data,processing data,establishing forest fire spreading model and simulating the spreading model with the modern survey technology,to provide data support for the forest fire police based on the established fire spreading model.This paper also try to explore new way for the study of forest fire spreading regulations.The background of this paper is the forest fire which took place in Gala Mountain of NenJiang county,HeiHe city of HeiLongJiang province in 2006,and the researching data is the high time-resolution MODIS remote sensing image.On the comparasion of different remote sensing data,the MODIS that has the anti-fire channels is choosen to be the data source of this researching .I get the MODIS remote sensing image dated from May 21^th to May 29^th.The remote sensing image has a spectrum from visible light to infrared and the resolution include 250 metre,500 metre and 1000 metre.I try to dine the forest fire spreading model based on those dynamic MODIS data.In the process of modeling,the first step is the pretreatment of remote sensing image.Then process the data fusion work for the image to add the fire information to the image.Next the image need to be vectorized using the ArcGIS software so as to get some data information which is necessary to the forest fire spreading model establishment.In the process of vectorizing, The determination of the gravity center of polygon of forest fire spreading is the first problem to be solved.The octree is imported into this process to make the work meet the custom requirement.On the basis of the characteristic of the data,the ER algorithm and LIR algorithm are used to ensure the precision of this forest fire spreading model.From the model I set up, the forecasting relative error which is smaller than 1% make up 50% of the whole relative error, while the relative error smaller than 5% make up 25%,and the error between 5% and 10% make up 50%. In order to test and verify the precision,the forest fire data from Kandu River,Miandu River and Russia are used and it turns out that the precision meet the model requirement.GM（1,N） model , which could provide a new way to reasearch the relatively forest fire factors.It is the fist time to put forward gray forest fire spreading model and CA model in the computer simulation and computer expression, then make the two model together .and achieve advantage complementary in it. at the same time, we also use ArcEngine to open up forest fire forecast system, which could successfully load remote sensing image, tracking the forest fire’s borderline ,modeling data and getting data automatically、 forecasting forest fire spreading borderline automatically, and could also make a compare for load remote sensing image.The main originality of this paper include:（1）Set up the forest fire spreading model using the remote sensing data for the first time.Thus this paper propose a new way for forecasting the spread of forest fire and establish a new technology system for forest fire spreading model establishment.（2）In the course of modeling,the method of robust estimation is imported and the ER algorithm and LIR algotirhm are proposed according to the characteristic of the modeling data.（3）The Gray Theory which is imported into the modeling research make the model simple,clear and visual with high precision.At the same time this model is in common use and easy to be implementde by computer to achieve auto-forecasting.（4）In the simulation on computer,the idea of improving the CA model on the basis of gray forest fire spreading model is proposed.Establish technic system for gray forest fire spreading model which base on MODIS data is really a complementarity to traditionally forest fire spreading model, after make a conclusion of past experience it make the data choosing and model theory together, in this way, we could see a bright future in establishing forest fire spreading model.更多还原