Key technologies of land use information extraction based upon multisource remote sensing data: a case study of hilly-plain transition region in the middle and lower reaches of the Yellow River
Key technologies of land use information extraction based upon multisource remote sensing data: a case study of hilly-plain transition region in the middle and lower reaches of the Yellow River
复制标题
DOI：
10.1117/12.869066
复制
发表时间：
2010-10
期刊：
影响因子：
--
通讯作者：
Heli Lu;Guifang Liu
中科院分区：
文献类型：
--
作者：
Heli Lu;Guifang Liu
利用GIS技术将多源遥感数据与非遥感数据相结合进行影像分类已成为一种趋势。本文利用多源遥感数据（TM和CBERS-02 B）、DEM、坡度数据、土地利用现状图和其他地理辅助数据，建立了土地利用信息提取的技术框架。结果表明：（1）联合收割机TM和CBERS-02 B数据具有相似的空间分辨率和光谱分辨率，可以联合作为土地利用源数据。本研究采用了多层次监督分类的方法。(2)利用GIS技术建立背景数据库，提高了解释精度。首先，整合地形图、土壤图、土地利用图、交通图等非遥感信息作为背景数据库;然后将土地利用分类与上述数据库进行重叠。结果表明，该方法的不确定度降低了23.2%. (3)研究区平原区、丘陵区和黄泛平原区的旱地谱差异很大，平原区和黄河漫滩区的居民地谱基本一致。针对上述“同物异谱”和“异物同谱”现象，建立了基于遥感图像与地理环境关系的专家知识库。因此，平均分类精度提高了12.1%。
It was a trend that image was classified through combining multi-source remote sensing data with non-remote sensing data by GIS technology. In this paper, technological framework of land use information extraction was established using multi-sources remote sensing data (TM and CBERS-02B), DEM, slope data, land use map and other geographic auxiliary data. The result showed :( 1) It was possible to combine TM and CBERS-02B as land use sources data because of their similar spatial resolution and spectral resolution. In this research the method of multi-level supervised classification was adopted. (2) Interpretation accuracy was improved by establishing background database through GIS technology. First, non-remote sensing information, such as topographic map, soil map, land-use map, transportation map, etc, was integrated as background database. Then land use classifications were overlapped with above database. The results showed that uncertainty could reduce by 23.2%. (3) In the study area dry land spectrums in plain area, hilly area and the Yellow River flood plain were very different and spectrums of habitation in plain and the Yellow River land wash were the same. As for above phenomenon of "same object with different spectrums" and "different objects with same spectrum", expert knowledge database was established based upon relationship between remote sensing image and geographical environment. As a result average classification accuracy was improved by 12.1%.