南京林业大学学报(自然科学版) ›› 2024, Vol. 48 ›› Issue (2): 19-26.doi: 10.12302/j.issn.1000-2006.202212016

• 专题报道Ⅰ: 山水林田湖草沙一体化保护———湿地生态研究专题(执行主编 张金池 薛建辉) • 上一篇    下一篇

基于Sentinel-1和Sentinel-2影像的洪泽湖国家湿地公园水生植被信息提取

韩森(), 阮仁宗*(), 傅巧妮, 许捍卫, 衡雪彪   

  1. 河海大学水文水资源学院,江苏 南京 210098
  • 收稿日期:2022-12-09 修回日期:2023-03-07 出版日期:2024-03-30 发布日期:2024-04-08
  • 通讯作者: *阮仁宗(ruanrenzong@163.com),副教授。
  • 作者简介:韩森(2902053779@qq.com)。
  • 基金资助:
    应急管理部国家减灾中心-中国人民财产保险股份有限公司联合实验室2020年度开放基金(NDRCCPICC202003)

Extraction of aquatic vegetation in Hongze Lake National Wetland Park based on Sentinel-1 and Sentinel-2 images

HAN Sen(), RUAN Renzong*(), FU Qiaoni, XU Hanwei, HENG Xuebiao   

  1. College of Hydrology and Water Resources, Hohai University, Nanjing 210098, China
  • Received:2022-12-09 Revised:2023-03-07 Online:2024-03-30 Published:2024-04-08

摘要:

【目的】 探索利用光学遥感和雷达遥感数据进行湖泊湿地水生植被信息提取方法。【方法】 以洪泽湖国家湿地公园为研究区,基于Sentinel-1的SAR影像和Sentinel-2的MSI影像,利用面向对象影像分析技术,结合EVSI、NDVI、SR特征指数和对象之间的上下文特征,以及挺水植被高度的差异所对应SAR影像上的后向散射系数,在对象级的基础上建立决策树模型对湿地水生植被进行分类,分析洪泽湖国家湿地公园水生植被以及挺水植被的分布状况。【结果】 研究区水生植被类群分类精度为89%,Kappa系数为0.85;挺水植被种群分类精度为85.2%,Kappa系数为0.76。与基于像元分析方法的结果相比,面向对象的影像分析方法具有更高的精度;湿地水生植被以沉水植被和挺水植被为主,其中挺水植被中以荷叶和芦苇为主。【结论】 本研究提出的湖泊湿地水生植被信息提取方法具可行性,可为湿地管理与决策提供科学依据。

关键词: 水生植被, Sentinel-1, Sentinel-2, 决策树, 植被特征指数, 后向散射系数, 洪泽湖国家湿地公园

Abstract:

【Objective】 The objective of this study was to explore the extraction of spatio-temporal distribution of aquatic vegetation in lake wetlands using Sentinel-1 and Sentinel-2 data. 【Method】 Hongze Lake National Wetland Park was chosen as the research area. Based on the combination of Sentinel-2 MSI images and Sentinel-1 SAR images, the object-oriented image analysis was used. The feature set was constructed by using EVSI, NDVI, SR feature index and contextual features between objects, as well as differences in the backscatter coefficients of the SAR images corresponding to differences in the height of the emergent vegetation types. A decision-tree model was established at the object level to classify the wetland, and the spatio-temporal distribution of the aquatic vegetation and the emergent vegetation in the Hongze Lake National Wetland Park was acquired. 【Result】 The classification accuracy and the Kappa coefficient of aquatic vegetation were observed to be 89% and 0.85, respectively, and that of the emergent vegetation was 85.2% and 0.76, respectively. The results showed that, compared with the results of the pixel-based analysis method, the accuracy of object-based image analysis was higher. The wetland aquatic vegetation was dominated by submerged and emergent vegetation; among the emergent vegetation, lotus leaves and reeds were dominant. 【Conclusion】 The methods proposed in this study were feasible, and the results could provide a scientific basis for managers and planners of wetlands.

Key words: aquatic vegetation, Sentinel-1, Sentinel-2, decision tree, vegetation characteristic index, backscatter coefficient, Hongze Lake National Wetland Park

中图分类号: