收藏设为首页 广告服务联系我们在线留言

干旱区地理 ›› 2019, Vol. 42 ›› Issue (4): 893-901.doi: 10.12118/j.issn.1000-6060.2019.04.21

• 地球信息科学 • 上一篇    下一篇

基于Landsat8影像时间序列NDVI的作物种植结构提取

白燕英1,高聚林2,张宝林3   

  1. 1内蒙古农业大学水利与土木建筑工程学院,内蒙古呼和浩特010018;2内蒙古农业大学农学院,内蒙古呼和浩特010019;3内蒙古师范大学化学与环境科学学院,内蒙古呼和浩特010018
  • 收稿日期:2019-01-12 修回日期:2019-04-13 出版日期:2019-07-25 发布日期:2019-07-25
  • 通讯作者: 高聚林(1964-),男,教授,博士生导师,主要从事作物生理生态及决策系统研究. E-mail:nmgaojulin@163.com
  • 作者简介:白燕英(1980-),女,副教授,博士,主要从事农业遥感研究. E-mail:baiyanying80@163.com
  • 基金资助:
    国家自然科学基金项目(51769023);华北黄土高原地区作物栽培科学观测实验站(25204120);国家重点研发计划项目(2017YFD0300800)

Extraction of crop planting structure based on timeseries NDVI  of Landsat8 images

BAI Yanying1,GAO Julin2,ZHANG Baolin3   

  1. 1College of Water Conservancy and Civil Engineering,Inner Mongolia Agricultural University,Hohhot 010018,Inner Mongolia,China;2Agricultural College,Inner Mongolia Agricultural University,Hohhot 010018,Inner Mongolia,China; 3College of Chemistry and Environmental Science,Inner Mongolia Normal University,Hohhot 010018,Inner Mongolia,China
  • Received:2019-01-12 Revised:2019-04-13 Online:2019-07-25 Published:2019-07-25

摘要: 为提高内蒙古平原灌区作物种植结构遥感监测精度和效率,提出一种基于时序NDVI曲线的作物种植结构提取方法。以内蒙古土默特右旗平原区为研究区域,以2015年覆盖作物生育期的多时相Landsat影像为数据源,根据不同地物其NDVI值范围不同,将研究区地表分为植被覆盖地表,无植被覆盖地表和水体3类。在植被覆盖区域内,根据林地和荒草地时序NDVI曲线特征,提取林地和荒草地,其余区域即为农田。根据小麦、玉米、葵花和西葫芦的时间序列NDVI曲线特征差异构建分类决策树模型,在农田区域内提取上述作物的空间种植分布信息。研究区各类地物及作物遥感提取面积与实际统计面积接近,土地利用分类总体精度达到85.71%,作物分类总体精度达到82.69%。研究结果表明该方法提取作物种植信息的精度较高,能够实现区域作物种植信息的高效准确监测。

关键词: Landsat, 时间序列, NDVI, 作物种植结构, 决策树

Abstract: In order to improve the accuracy and efficiency of remote sensing monitoring of crop planting information in plain irrigation area in Inner Mongolia,China,this paper proposes a method of crop planting information extraction based on timeseries NDVI profile.Selecting Tumoteyou County of Inner Mongolia as the study area,using the multitemporal Landsat images covering crop growing season in 2015 as data sources,the study area was divided into vegetation covering area,nonvegetation covering area and water covering area according to the NDVI value range of ground objects.In the area of vegetation covering,the forest area and grass area were extracted successively according to the NDVI curve differences of forest and grass.The rest areas were farmland in vegetation covering area.According to the NDVI profile of wheat,corn,sunflower and zucchini,the decisiontree was built,and the spatial planting information of the crops was extracted successively according to the decisiontree.Building the decisiontree is mainly based on the difference of NDVI average curve characteristics between crops,forest and grass.However,the change range of each crop NDVI due to their own different growth at the same time should also be fully considered.Appling the decisiontree,the extraction area of land use and crop is close to the actual statistical area.The overall accuracy of land use classification was 85.71% and the Kappa coefficient is 0.75.The overall accuracy of crop classification was 82.69% and the Kappa coefficient was 0.77.The results show that the research method can monitor crop spatial planting information in the region accurately and effectively.

Key words: Landsat, timeseries, NDVI, crop planting structure, decisiontree