摘要
利用简译遥感信息提取软件,以资源三号高分辨率卫星影像为数据源,选取老挝首都万象市为研究区,采用分层信息提取思想,在道路、建筑等地表覆盖要素人工提取结果的基础上,对未分类地物(水田、旱地、林地、工矿用地),利用最小距离算法和深度学习算法进行信息提取研究。研究结果表明,基于最小距离算法的总体分类精度为85.7%,Kappa系数为0.8165;基于深度学习算法的总体分类精度为89.29%,Kappa系数为0.8634,都可以达到很好的智能化提取效果。
Based on the results of manual extraction of road,building and other land cover elements,this paper uses the minimum distance method and depth learning method to classify the unclassified land(paddy field,dry land,forest land,industrial and mining land)by using the simple translation information extraction software,taking the high-resolution image No.3 as the data source and Vientiane City as the research area carry on the research of classification and extraction.The research results show that the overall classification accuracy based on the minimum distance algorithm is 85.7%,Kappa coefficient is 0.8165;the overall classification accuracy based on the deep learning algorithm is 89.29%,Kappa coefficient is 0.8634,which can achieve good intelligent extraction effect.
作者
杨昊
杨壮
孙迪
YANG Hao;YANG Zhuang;SUN Di(Heilongjiang Institute of Geomatics Engineering,Harbin 150081,China)
出处
《测绘与空间地理信息》
2021年第S01期168-170,176,共4页
Geomatics & Spatial Information Technology
关键词
资源三号
面向对象
卫星影像
信息提取
分割尺度
ZY-3
object-oriented
satellite image
information extraction
segmentation scale
