湖北农业科学 ›› 2022, Vol. 61 ›› Issue (16): 175-181.doi: 10.14088/j.cnki.issn0439-8114.2022.16.034

• 信息工程 • 上一篇    下一篇

基于Sentinel-2时序影像的水稻种植信息提取

汪荃1, 陈军军2   

  1. 1.中南勘察基础工程有限公司,武汉 430081;
    2.湖北正纽地理信息有限公司,湖北 黄石 435006
  • 收稿日期:2021-09-18 出版日期:2022-08-25 发布日期:2022-09-14
  • 通讯作者: 陈军军,主要从事地理信息工作。
  • 作者简介:汪 荃(1974-),男,湖北钟祥人,高级工程师,主要从事地理信息工作,(电话)18872302228(电子信箱)4373058@qq.com。

Extracting rice planting information based on Sentinel-2 time series images

WANG Quan1, CHEN Jun-jun2   

  1. 1. Central South Exploration & Foundation Engineering Co.,Ltd.,Wuhan 430081,China;
    2. Hubei Zhengniu Geographic Information Co.,Ltd.,Huangshi 435006,Hubei, China
  • Received:2021-09-18 Online:2022-08-25 Published:2022-09-14

摘要: 采用安徽省芜湖市芜湖县南部水稻种植区2019年3月至10月的Sentinel-2时序遥感影像,利用基于像素分类的支持向量机法、最大似然法,基于3种植被归一化植被指数(NDVI),比值植被指数(RVI)和归一化差异绿度指数(NDGI)的组合分类方法提取水稻种植信息,并结合目视解译结果对各种分类方法得到的分类结果依据混淆矩阵进行精度评价。结果表明,Sentinel-2遥感影像能够快速有效提取研究区域的水稻种植信息,其中最大似然法比支持向量机法更适合提取水稻信息,并且多时相影像数据的使用和相关植被指数的采用能够明显提升水稻信息提取精度,其最佳组合的水稻总体精度高达95.5%,Kappa系数达到了0.922,可作为水稻资源调查方法的一种有效补充手段。

关键词: Sentinel-2, 支持向量机法, 最大似然法, 时间序列, 植被指数

Abstract: Sentinel-2 time series remote sensing images from March to October 2019 in the southern rice planting area of Wuhu County, Wuhu City, Anhui Province were used, and the rice planting information was extracted by using the support vector machine method, the maximum likelihood method based on pixel classification and the classification method based on the combination of 3 Normalized Difference Vegetation Index (NDVI), Ratio Vegetation Index (RVI) and Normalized Difference Greenness Index (NDGI). The results showed that sentinel-2 remote sensing image could quickly and effectively extract the rice planting information in the study area, and the maximum likelihood method was more suitable for extracting rice information than the support vector machine method. The use of multi temporal image data and related vegetation index could significantly improve the accuracy of rice information extraction. The overall accuracy of the best combination of rice was as high as 95.5%, and the kappa coefficient was 0.922, which could be used as an effective supplementary method for rice resources investigation.

Key words: Sentinel-2, support vector machine classification, maximum likelihood classification, time series, vegetation index

中图分类号: