引用本文:杨玉婷,董秀春,刘忠友,蒋怡,李宗南,刘泳伶.基于Sentinel-2时序NDVI的麦冬识别研究[J].中国农业信息,2021,(3):35-42
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 447次   下载 260 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于Sentinel-2时序NDVI的麦冬识别研究
杨玉婷,董秀春,刘忠友,蒋怡,李宗南,刘泳伶
四川省农业科学院遥感与数字农业研究所,成都 610066
摘要:
【目的 】 为掌握四川省重要中草药麦冬的种植面积,开展麦冬遥感识别及空间信息提取研究。 【方法 】 文章以四川省绵阳市三台县麦冬主要种植区为研究区,选取2020年11月至2021年5月共计6期Sentinel-2遥感影像,结合实地调查数据建立麦冬、油菜、小麦3种地类的样本数据集,分析该时期内麦冬、油菜、小麦的NDVI时序差异,基于随机森林算法构建麦冬提取模型。随后选取区分度最大的3期NDVI作为输入变量与以全6期NDVI作为输入变量进行麦冬提取精度对比。 【结果 】 (1) 麦冬与油菜、小麦的NDVI在11月、3月、5月差异较大;(2) 以6期的NDVI作为输入变量的麦冬种植信息提取总体精度为91.92%,Kappa系数为0.892;(3) 以3期的NDVI作为输入变量的麦冬种植信息提取总体精度为90.05%,Kappa系数为0.823 2,分类精度略低于6期NDVI全输入,但基于3期关键节点的NDVI时序数据能较准确提取麦冬种植信息。 【结论 】 该结果可为四川省麦冬遥感识别和种植区变化监测提供参考。
关键词:  麦冬识别  时序NDVI  Sentinel-2  随机森林算法  中草药
DOI:10.12105/j.issn.1672-0423.20210304
分类号:
基金项目:作物信息多源遥感智能提取及农业空间大数据挖掘研究 (2021XKJS077);基于深度强化学习的柑橘农业大脑研发(2021XKJS076);遥感大数据与专家经验支持下的作物水肥智能决策技术和系统--以蔬菜为例(2021XKJS078);攀西地区紫茎泽兰遥感监测与生态风险评价研究(2020JDRC0129)
Study on Ophiopogon japonicus identification based on Sentinel-2 NDVI time series
Yang Yuting, Dong Xiuchun, Liu Zhongyou, Jiang Yi, Li Zongnan, Liu Yongling
Institute of Remote Sensing and Digital Agriculture,Sichuan Academy of Agricultural Sciences,Chengdu 610066,China
Abstract:
[ Purpose ] In order to grasp the planting area of ophiopogon japonicus,an important Chinese herbal medicine in Sichuan,study on rapid identification of ophiopogon japonicus and spatial information extraction based on remote sensing technology was carried out.[Method ] Taking the main planting area of ophiopogon japonicus in Santai County,Mianyang City,Sichuan Province as the study area,and the 6 images of Sentinel-2 from November 2020 to May 2021 were selected. Combining the field survey data to establish sample data sets of ophiopogon japonicus,oilseed rape and wheat three types of land cover,the study analyzed the NDVI time series differences between wheat,oilseed rape and ophiopogon japonicus during this period,and then constructed the ophiopogon japonicus extraction model based on the Random Forest algorithm. Subsequently,the three NDVI with the largest degree of discrimination was selected as the input variable and the whole six NDVI was used as the input variable to compare the extraction accuracy of ophiopogon japonicus.[Result ] The results are as follows:(1)The NDVI of ophiopogon japonicus,oilseed rape,and wheat are quite different in November,March,and May;(2)The overall accuracy of ophiopogon japonicus extraction with the six NDVI is 91. 92%,The Kappa coefficient is 0.892 0;(3)The overall accuracy of ophiopogon japonicus planting information extraction with the three NDVI is 90. 05%,the Kappa coefficient is 0.823 2,and the classification accuracy is slightly lower than the full input of 6 NDVI,but based on the key nodes of three NDVI time series data can more accurately extract the ophiopogon japonicus.[Conclusion ] The results can provide references for remote sensing identification of Ophiopogon japonicus and monitoring the change of planting areas in Sichuan.
Key words:  ophiopogon japonicus identification  NDVI time series  Sentinel-2  random forest algorithm  Chinese herbal medicine