基于决策树的港口城市遥感自动分类方法研究

摘要遥感影像自动分类一直是研究人员追求的目标，研究人员进行了大量的研究，提出了很多智能型算法、取得了许多的成绩，但是仍然存在一些问题和挑战。主要表现为：分类精度难以保证、自动化程度不高、专家特征知识融入困难等。本文在研究现有分类算法的基础上，针对面对极为复杂的数据处理地物的空间分布特征时传统分类方法的效果不佳这一问题，而决策树分类模型比较能够很好的解决这一问题，提出基于决策树的港口城市遥感自动分类方法。选取江苏省连云港市为重点研究区，以Landsat7 ETM+遥感影像为数据源，提取、分析研究区范围内典型地物的光谱特征信息，利用ENVI软件建立一组分类规则，构建适应于区域特征影像的决策树分类模型，实现了2008年连云港市的土地利用分类。并将实验结果与传统的最大似然分类法分类精度进行了对比分析。实验结果表明：采用本文的决策树分类法能够实现遥感影像的自动分类，总体精度达到97.2%，分类优于传统的最大似然法，基本能够满足行业应用的要求。此外，通过决策树规则集的构建，有利于专家知识的融入，分类精度可控。43632

毕业论文关键词：自动分类；遥感影像；决策树；规则集

Method for Remote sensing Image Automatic Classification based the Decision Tree

Abstract

Automatic remote sensing image classification has always been the target of the researchers, the researchers conducted a great deal of research, put forward a lot of intelligent algorithm, obtained many achievements, but there are still some problems and challenges. Main show is: the classification precision is difficult to guarantee, automation degree is not high, the characteristics of expert knowledge into difficulties.Based on the research on the basis of the existing classification algorithm, for in the face of extremely complicated data processing features of spatial distribution characteristics of the traditional classification method of poor effect of this problem, and the decision tree classification model to compare well to solve this problem, put forward the port city of remote sensing automatic classification method based on decision tree. Selection of lianyungang city, jiangsu province as the key research areas and Landsat7 ETM + remote sensing image as data source, extraction, analysis and research area within the scope of typical ground objects spectrum characteristic information, using the ENVI software to establish a set of classification rules, to construct the decision tree classification model of regional characteristics in image, to implement the land use classification in lianyungang city in 2008. And the experimental results with the traditional maximum likelihood classification classification accuracy is analyzed.The experimental results show that: the decision tree classification in this paper can realize the automatic classification of remote sensing image, the overall accuracy of 85%, superior to the traditional maximum likelihood classification method, basic can satisfy the requirements of the industry application. In addition, by decision tree to construct a rule set, is conducive to the integration of expert knowledge, classification accuracy control.

Key words: Automatic Classification Remote Sensing Image Decision Tree RuleSet

摘要 I

Abstract II

目录 III

1绪论 1

2决策树分类方法概述 2

2.1决策树的定义及其分类原理 2

3研究区及数据来源 3

3.1研究区概况基于决策树的港口城市遥感自动分类方法研究:http://www.youerw.com/guanli/lunwen_44556.html