摘要
为了及时快速地得到城市中水体、植被、不透水面和裸土四类城市典型地物的覆盖信息,建立了一种使用归一化植被指数、改进型水体指数以及基于光谱分析后新建立的比例指数,结合伪彩色合成分类的方法。通过对分类地物的光谱特征分析,综合使用比例指数和改进型水体指数对不透水层信息进行了提取。验证结果表明,该方法仅需要通过颜色特征就可快速生成四类地物信息分布图,无需任何分类器算法。通过与同时相高分辨率影像中的200个随机点信息进行比对后,得到该方法可达到83%的总体分类精度,Kappa系数0.803,四类地物的用户精度和制图精度均达到80%以上。
To get the typical land cover information of water, vegetation, impervious surface and soil in an urban area timely, a method is presented using false color composition to get classification information, which is based on normalized difference vege- tation index （NDVI）, normalized difference water index （MNDWI）, and ratio index （RI） we build based on spectral feature analysis are modified. In impervious surface extraction, MNDWI and RI are used to get it. The result shows that the method presented is able to get the map of four typical land covers in an urban area quickly without any algorithms. Two hundred sample points are randomly selected for an accuracy assessment using high resolution image obtained at almost the same time as refe- rence, the overall accuracy of the classification is 83 % with the Kappa coefficient of 0. 803, and the user accuracy and map accu racy all can reach above 80%.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第3期1088-1094,共7页
Computer Engineering and Design
基金
国际科技合作基金项目(2012DFA10680
2010***)
2012山西省专业创新科技基础平台基金项目(2012091014)
关键词
遥感图像处理
模式识别
不透水层信息提取
遥感图像分类
伪彩色
remote sensing image processing, pattern recognition, impervious surface informationt extraction, remote sensingimage interpretation, false color
