摘 要植被不仅是环境重要的组成部分，而且是反映地方生态环境的最好标志之 一。植被研究是遥感的重要应用领域。因此，对遥感影像上的植被信息进行提取 这项工作具有重要的实际应用价值。近年来，国产高分辨率遥感卫星得到了迅猛 发展。它所具备的不断提高的影像空间分辨率、渐渐增强的影像获取能力、较好 的影像现势性等特点开始打破国外商业卫星主导的局面，并逐步广泛服务于各行 业用户，其中高分辨率对地观测系统（高分系列）在植被遥感中愈来愈受到关注。69969

本研究以徐州市为研究区。采用国产高分一号（GF-1）高分辨率数据，进 行了图像辐射定标、几何纠正、大气校正、裁剪研究区等预处理。通过监督分类 的最大似然算法、非监督分类下的迭代自组织数据分析算法（ISODATA）和决 策树分类算法分别对徐州市植被信息进行提取。结果表明，针对徐州的地表特点， 最大似然分类法提取徐州植被信息是最为适宜的。在最大似然分类中，总体分类 精度为 97.45%，Kappa 系数为 92.56%。在决策树分类中，总体分类精度为 83.67%， Kappa 系数为 74.72%。ISODATA 分类法中，因为不需要设置训练样本，所以分 类精度不高，所以此方法在本研究中并不太适用。

毕业论文关键词：遥感 植被信息提取 徐州 高分一号 分类

Research on Vegetation Information Extraction in Xuzhou City Based on the Data of GF-1

Abstract

Vegetation is not only an important part of the environment,and is one of the best marks to reflect the local ecological environment.Vegetation research is an important application field of remote sensing. Therefore, the work of extracting the vegetation on the remote sensing image information  has important practical application   value. In recent years, China-made high-resolution remote sensing satellites have been developing rapidly. It has continued to improve the spatial resolution  of the image, and gradually enhanced image capture capabilities, a better image of the current situation and other characteristics.Such situation now begins to break down foreign commercial satellite dominant situation, and be widely used in various industries. High resolution earth observation system (GF-1) in vegetation remote sensing becomes more and more attention.

The study is based on Xuzhou City area. It uses domestic GF-1 high-resolution data and does pretreatment, such as image radiation correction,geometric correction,atmospheric correction,cutting the study area.Through the maximum likelihood algorithm of supervised classification and unsupervised classification of ISODATA classification and decision tree classification algorithm in xuzhou city vegetation information was extracted respectively.The results showed that, for Xuzhou surface characteristics, the decision tree classification method to extract information Xuzhou vegetation is the most suitable.In the maximum likelihood classification, the overall classification accuracy is 97.45%, the Kappa coefficient is 92.56%. In the decision tree classification, the overall classification accuracy is 83.67%, the Kappa coefficient is 74.72%.In the ISODATA classification,it does not need to set the training sample,the classification accuracy is not high, so this  method in the study is not applicable.

Key Words: Remote sensing Vegetation information extraction Xuzhou GF-1 Classification

目 录

摘 要