“高分五号”高光谱互花米草遥感识别与制图研究  

Monitoring the Invasion of Spartina Alterniflora Using Hyperspectral Remote Sensing Image of GF-5

作　　者：任广波 周莉 梁建 路峰 王安东 王建步 李晓敏 马毅 REN Guang-bo;ZHOU Li;LIANG Jian;LU Feng;WANG An-dong;WANG Jian-bu;LI Xiao-min;MA Yi(First Institute of Oceanography,MNR,Qingdao 266061,China;Shandong Yellow River Delta National Nature Reserve Administration,Dongying 257091,China;Qidong Photoelectric Remote Sensing Center,Shanghai Institute of Technical Physics,CAS,Nantong 226200,China)

机构地区：[1]自然资源部第一海洋研究所,山东青岛266061 [2]山东黄河三角洲国家级自然保护区管理委员会,山东东营257091 [3]中国科学院上海技术物理研究所启东光电遥感中心,江苏南通226200

出　　处：《海洋科学进展》2021年第2期312-326,共15页Advances in Marine Science

基　　金：国家自然科学基金项目——海岸带遥感影像半监督学习自动化分类方法研究—以黄河三角洲滨海湿地分类为例(41206172);国家重点基础研究发展计划项目——北极环境遥感与数值预报合作平台建设(2018YFC1407200)。

摘　　要：基于高分五号卫星上搭载的世界上首个同时具有大刈幅、宽谱段的高光谱遥感载荷获取的遥感图像,以黄河三角洲滨海湿地为研究区,开展了对外来入侵物种互花米草的分类和制图。通过对比代表不同分类思想的SVM,SCM,ANN和MLC方法在全波段参与分类和特征波段分类中的结果,发现SVM分类方法在不同的数据中均能得到最好的分类效果,总体分类精度达到了94.23%。虽然利用特征波段的分类结果总体精度较全波段低(约3%),但更节省计算资源和计算时间。根据分类结果,在研究区内互花米草的分布面积达到了3863.16 hm^(2),占调查区域潮间带总面积的21.23%,但却占据了海岸线总长度的51.47%,其中121平台以北海岸线被100%占据,而该平台以南的海岸却几乎没有互花米草。结合现场调查数据和NDVI等植被指数的分析结果,发现互花米草的生长状态好坏与其淤泥底质的形成时间点相关,即越新形成的淤泥质潮滩上互花米草的长势越好,且好、中、差三种类型互花米草的面积比率是7∶4∶3,表明了该区域的互花米草仍有较为旺盛的生命力和入侵扩展能力。The invasive species Spartina alterniflora in the coastal wetlands of the Yellow River Delta were classified and mapped using a hyperspectral remote sensing image from the newly launched Chinese satellite"GF-5"with a wide coverage and wide spectral band at the same time.By comparing the results of SVM,SCM,ANN and MLC,which represent the mostly adopted typical classification framework,we find that SVM can achieve the best classification results in both full-bands data and dimension-decreased feature-bands data,and the overall classification accuracy reaches 94.23%.Although the overall accuracy of feature-bands data classification result is lower than that of full-bands classification(about 3 percent),we believe that a little loss of accuracy is worthwhile in order to save computing resources and processing time.It is concluded that the spectral heterogeneity of Spartina alterniflora and the low spatial resolution of the used remote sensing data are the main reasons for the misclassification of Spartina alterniflora with other coastal plant such as Phragmites australis.According to the classification and mapping results,the distribution area of S.alterniflora in the study area reached 3863.16 ha,accounting for 21.23%of the total intertidal zone area.51.47%of the total length of the coastline(sea water boundary)is covered by the Spartina alterniflora.Among them,the coastline in the north of the"121 oil-drilling platform"was 100%covered by Spartina alterniflora,while the southern part was almost empty.Based on field survey data and NDVI analysis,it is found that the growth status of Spartina alterniflora is related to the formation time of muddle flat,that is,the S.alterniflora grows better on the newly formed silty tidal flat.The area ratio of three types of Spartina alterniflora was 7∶4∶3,which indicates that Spartina alterniflora still has strong vitality,and invasion and expansion ability.

关 键 词：高分五号卫星 高光谱遥感 入侵物种监测 黄河三角洲 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]