基于蚁群优化的遥感影像分类研究
发布时间：2019-03-21 08:26
【摘要】：本文研究的方向是基于蚁群优化数据分类规则挖掘的遥感影像分类,总体来讲是从三个方面讨论的,首先是遥感影像分类问题的一般方法和特点；再者是蚁群优化的基本原理和数学模型以及实现的算法流程；最后是蚁群优化应用于遥感影像分类所要解决的问题。 关于遥感影像分类,使用了最新的Landsat-8数据,讨论了纹理变换、植被变换、主成分分析、独立成分分析、地形因子提取和多波段最佳指数因子提取等问题。把包含光谱、遥感指数、纹理、地形和线性变换的特征组合成一个多波段文件,作为分类的初始特征集,给出自己的特征选取方案。 关于蚁群优化,把双桥实验作为引例,讨论了旅行商问题、数据挖掘的蚁群优化原理、数学模型和算法流程,重点介绍了Ant-Miner模型,在总结蚁群优化一般框架的基础上,引入新的改进方案。 把遥感影像分类和蚁群优化结合来讲,具体讨论了数据离散、规则构造、规则剪枝、信息素策略和启发策略等方面问题的解决办法。最后设计实例比较基于蚁群优化分类方法与最大似然方法的分类结果,得出结论。
[Abstract]:The research direction of this paper is remote sensing image classification based on ant colony optimization data classification rule mining. Generally speaking, it is discussed from three aspects. Firstly, the general methods and characteristics of remote sensing image classification are discussed. Thirdly, the basic principle and mathematical model of ant colony optimization as well as the algorithm flow are discussed. Finally, the problem that ant colony optimization is applied to remote sensing image classification is solved. As for the classification of remote sensing images, the latest Landsat-8 data are used. The problems such as texture transformation, vegetation transformation, principal component analysis, independent component analysis, terrain factor extraction and multi-band optimal index factor extraction are discussed. The features including spectrum, remote sensing index, texture, terrain and linear transformation are combined into a multi-band file as the initial feature set of classification, and their feature selection scheme is given. Regarding ant colony optimization, taking double-bridge experiment as an example, this paper discusses traveling salesman problem, ant colony optimization principle of data mining, mathematical model and algorithm flow, and mainly introduces Ant-Miner model, on the basis of summarizing the general frame of ant colony optimization, Introduce a new improved scheme. Combining remote sensing image classification with ant colony optimization, the solutions of data discretization, rule construction, regular pruning, pheromone strategy and heuristic strategy are discussed in detail. Finally, an example is given to compare the classification results based on ant colony optimization method and maximum likelihood method, and the conclusion is drawn.
【学位授予单位】：安徽理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751;TP18
本文编号：2444763
[Abstract]:The research direction of this paper is remote sensing image classification based on ant colony optimization data classification rule mining. Generally speaking, it is discussed from three aspects. Firstly, the general methods and characteristics of remote sensing image classification are discussed. Thirdly, the basic principle and mathematical model of ant colony optimization as well as the algorithm flow are discussed. Finally, the problem that ant colony optimization is applied to remote sensing image classification is solved. As for the classification of remote sensing images, the latest Landsat-8 data are used. The problems such as texture transformation, vegetation transformation, principal component analysis, independent component analysis, terrain factor extraction and multi-band optimal index factor extraction are discussed. The features including spectrum, remote sensing index, texture, terrain and linear transformation are combined into a multi-band file as the initial feature set of classification, and their feature selection scheme is given. Regarding ant colony optimization, taking double-bridge experiment as an example, this paper discusses traveling salesman problem, ant colony optimization principle of data mining, mathematical model and algorithm flow, and mainly introduces Ant-Miner model, on the basis of summarizing the general frame of ant colony optimization, Introduce a new improved scheme. Combining remote sensing image classification with ant colony optimization, the solutions of data discretization, rule construction, regular pruning, pheromone strategy and heuristic strategy are discussed in detail. Finally, an example is given to compare the classification results based on ant colony optimization method and maximum likelihood method, and the conclusion is drawn.
【学位授予单位】：安徽理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751;TP18
【参考文献】
相关期刊论文 前10条
1 琚存勇;蔡体久;冯仲科;;基于数学形态学和最大似然法的遥感图像分类研究[J];北京林业大学学报;2005年S2期
2 李百寿;秦其明;许军强;张万峰;姚云军;;遥感图像线性影纹理解专家系统设计与实现[J];测绘科学;2008年02期
3 黄明;吴延斌;;基于混沌遗传算法的遥感影像分类[J];测绘科学;2011年02期
4 戴芹;刘建波;;基于蚁群规则挖掘算法的多特征遥感数据分类[J];地理研究;2009年04期
5 肖鹏峰,刘顺喜,冯学智,赵书河,赵萍;中分辨率遥感图像土地利用与覆被分类的方法及精度评价[J];国土资源遥感;2004年04期
6 贾坤;李强子;田亦陈;吴炳方;;遥感影像分类方法研究进展[J];光谱学与光谱分析;2011年10期
7 曹敏;史照良;阳建逸;;蜂群智能算法的遥感影像分类方法[J];测绘学报;2013年05期
8 谢宏,程浩忠,牛东晓;基于信息熵的粗糙集连续属性离散化算法[J];计算机学报;2005年09期
9 詹士昌,徐婕,吴俊;蚁群算法中有关算法参数的最优选择[J];科技通报;2003年05期
10 宋传中,朱光,刘国生,牛漫兰;淮南煤田的构造厘定及动力学控制[J];煤田地质与勘探;2005年01期
相关博士学位论文 前1条
1 周爽;蚁群算法在高光谱图像降维和分类中的应用研究[D];哈尔滨工业大学;2010年
，本文编号：2444763
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2444763.html
