文献类型：期刊文献

中文题名：基于高分一号时间序列数据的沙化土地分类

英文题名：Sandy lands classification using GF-1 time series NDVI data

作者：丁相元[1] 高志海[1] 孙斌[1] 吴俊君[2] 薛传平[1] 王燕[1]

第一作者：丁相元

机构：[1]中国林业科学研究院资源信息研究所;[2]中国科学院遥感与数字地球研究所

年份：2017

卷号：29

期号：3

起止页码：196-202

中文期刊名：国土资源遥感

外文期刊名：Remote Sensing for Land & Resources

收录：CSTPCD;;北大核心:【北大核心2014】；CSCD:【CSCD2017_2018】；

基金：国家高分辨率对地观测重大专项"高分林业遥感应用示范系统(一期)"(编号:21-Y30B05-9001-13/15)

语种：中文

中文关键词：GF-1影像;时间序列NDVI;沙化土地;应用潜力

外文关键词：GF-1 data;;time series NDVI;;sandy lands;;application potential

分类号：TP79

摘要：以高分一号(GF-1)16 m空间分辨率多光谱影像为数据源,对沙化土地类型的光谱特征以及其全年的NDVI变化特征进行了分析,发现时间序列数据变化信息可提高沙化土地类别之间的可分离度。对单一时相影像的分类结果和加入时间序列NDVI之后的分类结果进行了对比分析,结果表明,基于生长季单一时相原始影像的分类结果精度为73.34%,Kappa系数为0.7;非生长季单一影像与NDVI时间序列数据的分类结果总体精度为81.44%,Kappa系数为0.77;生长季单一时相影像并加入NDVI时间序列数据之后精度提高到了92.04%,Kappa系数达0.87,明显改善了对沙化土地类型的识别精度。表明单时相影像结合时间序列NDVI数据在沙化土地分类识别中有巨大的应用潜力。
In this study,GF-1 16 m multispectral images were used as data source,the spectral characteristics of each type of sandy land and its change characteristics of time series NDVI were analyzed,the sandy lands were classified by the GF-1 image at a single time,and time series NDVI data were compared with each other separately; on such a basis,the classification accuracy was evaluated. The results showed that the accuracy was73. 34% and Kappa coefficient was 0. 7 by only using single time original data in growing season; however,the accuracy was increased to 92. 04% by joining the time series NDVI data,with Kappa coefficient raised to 0. 87;the accuracy was 81. 44% and Kappa coefficient was 0. 77 by using the time series NDVI data combined with non-growing season data,thus improving the classification accuracy obviously. It is indicated that GF-1 time series NDVI data have a huge application potential in the sandy lands classification.