【作者】 胡全；

【导师】 王霓虹；

【作者基本信息】 东北林业大学 ， 森林工程， 2014， 博士

【摘要】 森林是人类宝贵的物质资源,在森林火灾发生时,为了尽快发现火情,及早扑灭火灾,做到“打早、打小、打了”,需要及时掌握火点周围基本情况,才能对火势发展进行准确预测和模拟。本文以伊春红星林业局二皮河林场为主要研究对象,针对目前GIS系统涉及的一些关键技术算法进行了深入的研究,提出了新的改进算法和实现途径,推动了森林火场GIS系统的深入发展。本文主要研究了以下4个方面内容：(1)基于多特征相融合和支持向量机的森林火场识别技术针对森林火场的快速识别采用了新的颜色特征提取方法,融合图像的颜色、纹理和形状特征作为图像的特征向量,并用支持向量机作为学习工具。利用已有森林火场的图像数据进行学习,提高森林火场自动识别的准确率。实验结果表明,基于HSV空间的18色非均匀量化算法,与传统166色和36色算法相比,特征维数和计算量都减少了,但检索效率却提升了0.6%,适用于森林火场的快速识别。设计编写了森林火点图像识别系统,此系统具备基本的特征提取、融合、标注、识别、SVM和数据维护等功能,在应用中取得了较好的实施效果。(2)森林火场三维地形简化算法深入研究了当前主流的三维地形简化算法,并对这些地形简化算法进行分析比较。对基于视点的混合结构四叉树地形简化算法进行了改进,构建了基于四叉树结构的多分辨率纹理模型和基于四叉树结构的空间索引,形成了一整套基于四叉树结构的森林火场三维地形数据建模和纹理映射的高效处理方案,使得森林火场的三维漫游速度提升了近2倍。(3)基于GPU技术的三维地形模拟结合前面的三维地形简化技术,使用Chunked LOD算法,实现了基于GPU技术的森林火场三维地形模拟,允分发挥了GPU的并行运算能力,与传统采用CPU计算方式相比,减少了在计算方面的投入费用,提高了森林火场三维地形的绘制效率。(4)林火蔓延模型及其算法改进重点研究了可燃物、地形坡度和气象条件等影响因子对森林火场的影响,通过对比分析目前常用的几种林火蔓延模型特点和适用范围,提出了一种基于王正非林火蔓延模型的改进算法,并引入元胞自动机技术,实现了基于元胞自动机的林火蔓延模拟,且在实验中加以验证。本文以伊春红星林业局二皮河林场为主要研究对象,提出了一整套以火点定位、三维地形简化和林火蔓延为主线的解决方案,有力的推动了我国森林火场模拟系统的发展,具有一定的研究和应用价值。更多还原

【Abstract】 Forest is a kind of valuable material resource. It is necessary to have a knowledge of the basic situation around the fire spot in order to discover the fire and put out the fire as soon as possible when forest fire happens, so that we can master it before hand and deal with it in time, and predict and stimulate the spread of the fire. This paper takes Erpi River Forestry Station of Hongxing Forestry Bureau in Yichun as the main research object, conducts in-depth research on key technology related in the GIS system, and puts forward new improved algorithm and its way of application, which promotes further development of GIS system in forest fire. This paper mainly states from the following four aspects:(l)Method Based on multi-features and Support Vector Machine for forest fire recognitionIt adopts new way of color feature extraction, taking the features like the color of the image, texture shape and so on as the eigenvector of the image, and taking Support Vector Machine as the tool of learning. Takes the advantage of existing image data, enhance the accuracy of automatic identification in forest fire. The research result shows that18-color non-uniform quantization algorithm which is based on HSV can be applied to quick recognition of fire ground. Compared with traditional166-color and36-color algorithm, the feature dimension and calculation goes down but the retrieval efficiency goes up by0.6percent. The author have designed a image system of forest fire recognition that include basic functions like feature extraction, integration, tagging, recognition, SVM and data maintenance and have a great effect in application.(2)Simplified algorithm of three-dimensional terrain in forest fireThis thesis is to study the simplified algorithm of three-dimensional terrain, and make a comparative analysis among these simplified algorithms. Besides, this paper improves the simplified algorithm for the mixed-structure quadtree terrain, and builds the multi-resolution texture model and spatial index both based on quadtree structure, which consist a set of data modeling of three-dimensional terrain of the forest fire based on quadtree structure and an efficient deal of texture mapping. It makes the speed of three-dimensional roaming increase by(3)Three dimensional terrain simulation based on GPUCombined with three-dimensional terrain technology, the paper uses ChunkedLOD to achieve the three dimensional terrain simulation which is based on GPU and. Compared with traditional CPU, the method makes full use of the parallel operation ability of GPU, reduces the cost in computing input and improves the rendering efficiency of three-dimensional terrain in forest fire. (4)Forest fire spreading model and improving of algorithmThe paper pays close attention to the factors which may influence fire ground like combustible, topographic slope and meteorological condition. By comparing the features and scope of application of several forest fire spreading model, this paper puts forward a algorithm based on Wang Zhengfei fire spreading model and introduces cellular automata technique into algorithm, taking simulation of forest fire which is based on cellular automata technique into practice.ErPi River Forestry Station of Hongxing Forestry Bureau in Yichun was the research example, and a set of solution of forest fire was proposed by employing dimensional GIS system which was consisted of the forest fire automatic positioning system, terrain simplification model structure, and fire spreading model. This solution would actively promote the development of forestry fire GIS system in our country and have a research and application value.更多还原