结合潮位与DEM的红树林遥感识别研究  

Identification of mangrove forest via remote sensing combined with tidal level and DEM

作　　者：张雪红 葛州徽 甄晓菊 姜楠 董天赐 ZHANG Xuehong;GE Zhouhui;ZHEN Xiaoju;JIANG Nan;DONG Tianci(School of Remote Sensing&Geomatics Engineering,Nanjing University of Information Science&Technology,Nanjing 210044,China;Hebei Provincial Key Lab for Meteorology and Eco-environment,Shijiazhuang 050021,China)

机构地区：[1]南京信息工程大学遥感与测绘工程学院,南京210044 [2]河北省气象与生态环境重点实验室,石家庄050021

出　　处：《南京信息工程大学学报》2024年第5期727-736,共10页Journal of Nanjing University of Information Science & Technology

基　　金：国家自然科学基金(41871239);中国气象局雄安大气边界层重点开放实验室(培育)开放课题(2023LABL-B16);2020年度江苏高校“青蓝工程”优秀青年骨干教师;河北省省级科技计划(21567624H)。

摘　　要：以广西北部湾为研究区,针对潮位周期性变化导致稀疏低矮红树林难以被准确提取的问题,基于多潮位Landsat8 OLI图像和数字高程模型(DEM)数据,通过构建红树林识别决策树模型,并以支持向量机(SVM)为对照,评价结合潮位和DEM信息的决策树法提取红树林信息的可行性.研究结果表明:1)不同高度、不同密度以及不同潮位红树林之间光谱差异均较大,稀疏低矮红树林也与阴坡林地、水体-陆生植被混合像元光谱存在严重“异物同谱”效应;2)无论是基于低潮位、高潮位图像,还是多潮位图像,相比未区分高度和密度,在SVM中将细分为高密红树林和稀矮红树林,其总体精度(分为红树林和非红树林两类)可分别提高4.65、4.41和7.22个百分点;3)基于多潮位图像及DEM的决策树模型识别的总体精度和Kappa系数分别为98.80%和0.973,比SVM中最佳值分别高出1.62个百分点和0.035.因此,通过同时考虑红树林高度、密度、潮位和DEM等特征,可明显提高红树林遥感识别的精度.To address the accurate extraction of sparse and low mangroves perplexed by the periodic change of tide level,we take the Beibu Gulf of Guangxi as the research area to construct a decision tree model for mangrove identi-fication using Landsat8 OLI images at low and high tidal levels and DEM(Digital Elevation Model)data,which is then evaluated by comparing with SVM(Support Vector Machine).The research results show that difference exists in the spectra of mangroves with different heights and canopy densities or under different tide levels,while the sparse and low mangroves share the same spectrum with shady slope forest and water-terrestrial vegetation mixed pixel.The SVM approach classifies the mangroves as high-dense type and low-sparse type,and improves the overall accuracy by 4.65,4.41 and 7.22 percentage points for low-tide,high-tide and multi-tide images,respectively.The proposed approach reaches 98.80%of overall accuracy and 0.973 of Kappa coefficient,which are 1.62 percentage points and 0.035 higher than the best values of SVM approach.It can be concluded that considering the mangrove height,density,tide level and DEM can significantly improve the identification accuracy of mangroves from remote sensing images.

关 键 词：红树林 Landsat 8 OLI 潮位 数字高程模型(DEM) 决策树 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置] S796[自动化与计算机技术—控制科学与工程]