摘要
采用先进遥感手段加强对滨海湿地资源的监测对其管理与保护至关重要。以江苏盐城国家级珍禽自然保护核心区为实验区,借助GEE云平台对获取的湿地植被生长关键期(5月至9月)的Sentinel-1 SAR和Sentinel-2多光谱时间序列影像基于中值法进行月影像合成,然后对5个单月份数据集和4个不同时间跨度的多月份数据集基于7种不同的波段组合方案构成63个测试集,最后基于随机森林方法对各测试集进行湿地分类制图与比较分析。结果表明,与仅利用Sentinel-2的可见光与近红外波段相比,增加短波红外波段能显著提高滨海湿地的分类精度,总体精度最高可提高8%以上,增加红边波段总体精度最高可提高5%;与仅使用Sentinel-2光谱波段相比,将常用植被指数和Sentinel-1 SAR数据添加到湿地分类中,总体分类精度最高达到了90.6%。实验也验证了使用多月份时间序列遥感数据合成的分类效果要优于单月影像分类结果。研究结果为区域尺度中/高分辨率遥感影像的湿地分类提供了一定的技术参考。
Using advanced remote sensing technology to classify and map coastal wetlands is vital for wetland environment management and protection.Taking the core area of Yancheng National Rare Bird Nature Reserve,Jiangsu,as a typical area,the study obtained Sentinel-1 and Sentinel-2 time-series images of the critical period of vegetation growth(May to September)to composite monthly images on GEE platform.A total of 63 test sets were formed of five single month datasets and four multi-month datasets with different time spans using 7 different band combinations.Wetland classification mapping and comparative analysis were performed on each test set using the random forest method.The results showed that,compared to those that used only Sentinel-2 data in the visible and NIR bands,adding the SWIR bands could significantly improve the classification results with an overall accuracy improvement over 8%.An overall accuracy improvement of about 5%was achieved by adding the red-edge band.Compared to those that used only Sentinel-2 spectral information,adding vegetation index and Sentinel-1 SAR information to wetland classification achieved the highest overall accuracy of 90.6%.Further,the experiments also verified that the classification effect of using multi-month time-series remote sensing data was better than that of using single month data.This research provided technical reference for wetland classification in regional scale with medium/high resolution remote sensing images.
作者
陈媛媛
郭莹莹
魏翀
向云飞
朱梦瑶
张昕
CHEN Yuan-yuan;GUO Ying-ying;WEI Chong;XIANG Yun-fei;ZHU Meng-yao;ZHANG Xin(College of Civil Engineering,Nanjing Forestry University,Nanjing 210037,China)
出处
《长江流域资源与环境》
CAS
CSCD
北大核心
2024年第8期1691-1701,共11页
Resources and Environment in the Yangtze Basin
基金
国家自然科学基金项目(42101430)
江苏省自然科学基金资助项目(BK20180779)
南京林业大学青年科技创新基金项目(CX2018015)
实景地理环境安徽省重点实验室开放基金项目(2022PGE006)。
