基于物候特征的连云港滨海湿地植被提取方法  

Study on Vegetation Extraction Method of Lianyungang Coastal Wetland Based on Phenological Characteristics

作　　者：黄唯澄 高祥伟 陶洋 王圳 高亚军 李子威 鞠海建 HUANG Weicheng;GAO Xiangwei;TAO Yang;WANG Zhen;GAO Yajun;LI Ziwei;JU Haijian(School of Marine Technology and Geomatics,Jiangsu Ocean University,Lianyungang 222005,China;Lianyungang Forestry Technical Guidance Station,Lianyungang 222002,China;Nantong Jianghai Institute of Surveying and Mapping Co.,Ltd.,Nantong 226000,China)

机构地区：[1]江苏海洋大学海洋技术与测绘学院,江苏连云港222005 [2]连云港市林业技术指导站,江苏连云港222002 [3]南通市江海测绘院有限公司,江苏南通226000

出　　处：《江苏海洋大学学报（自然科学版）》2024年第3期26-33,共8页Journal of Jiangsu Ocean University:Natural Science Edition

基　　金：自然资源部滨海盐沼湿地生态与资源重点实验室开放基金项目(KLCSMERMNR2021102);连云港市科技局项目(SF2240)。

摘　　要：滨海湿地是介于海陆间的特殊生态系统,滨海湿地植被对其生态系统的功能具有重要影响。快速、准确的滨海湿地植被提取方法对滨海湿地的生态保护和管理具有重要意义。滨海湿地植被类型多为草本植被,不同植被类型之间影像的光谱特征和空间特征相似,可分离度小,导致植被遥感分类难度较大,而融合植被物候特征成为提高分类精度的重要手段。以连云港滨海湿地为研究区,利用PIE-Engine遥感云计算平台,获取2022年72景Sentinel-2影像构建NDVI时间序列模型,运用傅里叶函数(Fourier)拟合植被物候特征曲线,分析植被物候特征,并融合物候特征进行植被分类。结果显示:融合物候特征后,植被分类总体精度为83.83%,Kappa系数为0.76,相较于单时相影像方法,分类精度提高了16.6百分点,Kappa系数提高了0.23。因此,利用植被物候特征能有效地提高分类精度。Coastal wetland is a special ecosystem between land and sea.The vegetation of coastal wetland has an important influence on the function of its ecosystem.It is of great significance to study the rapid and accurate extraction method of coastal wetland vegetation for the ecological protection and management of coastal wetland.The vegetation types of coastal wetlands are mostly herbaceous vegetation.The spectral characteristics and spatial characteristics of images between different vegetation types are similar,and the separability is small,which leads to the difficulty of vegetation remote sensing classification.Fusion of vegetation phenological characteristics has become an important means to improve classification accuracy.In this study,the coastal wetland of Lianyungang was taken as the research area,and the PIE-Engine remote sensing cloud computing platform was used to obtain 72 scenes of Sentinel-2 images in 2022 to construct NDVI time series model.The Fourier function was used to fit the vegetation phenological characteristic curve,analyze the vegetation phenological characteristics,and integrate the phenological characteristics for vegetation classification.The results show that the overall accuracy of vegetation classification is 83.83%and the Kappa coefficient is 0.76 after the fusion of phenological features.Compared with the single-phase image method,the classification accuracy is increased by 16.6 percentage points and the Kappa coefficient is increased by 0.23.Therefore the use of vegetation phenological features can effectively improve the classification accuracy.

关 键 词：滨海湿地植被 物候特征 Sentinel-2影像 PIE-Engine平台 时间序列 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置] TP391.41[自动化与计算机技术—控制科学与工程]