应用珠海一号卫星高光谱数据的红树林遥感提取方法  

Remote Sensing Extraction Method of Mangrove Forest Using Zhuhai-1 Satellite Hyperspectral Data

作　　者：边福强 王晓宇[1] 杨志[1] 范东栋[1] BIAN Fuqiang;WANG Xiaoyu;YANG Zhi;FAN Dongdong(DFH Satellite Co., Ltd., Beijing 100094, China)

机构地区：[1]航天东方红卫星有限公司,北京100094

出　　处：《航天器工程》2021年第6期182-187,共6页Spacecraft Engineering

摘　　要：利用珠海一号卫星高光谱数据,以广东省沿海地区作为红树林监测的研究区域,开展了辐射定标、大气校正等预处理,完成了归一化水体指数(NDWI)、归一化植被指数(NDVI)的计算。通过综合NDWI-NDVI指数阈值法完成对红树林掩膜范围的提取,并基于掩膜范围均匀构建样本点,通过最大似然法、神经网络法、支持向量机法和马氏距离法4种监督分类方法,分别对红树林开展了监督分类,利用混淆矩阵计算出了分类精度和卡帕系数。仿真结果表明:文章提出的方法体现出了较好的样本分离度,所构建的样本点能较好地反映红树林的光谱信息,并能满足多种监督分类方法的需求,实现高精度红树林遥感提取与分析。Based on Zhuhai-1 satellite hyperspectral data,the coastal area of Guangdong province is used as the research area for mangrove forest monitoring,and preprocessing efforts such as radiometric calibration and atmospheric correction are carried out to complete the calculation.NDWI(normalized difference water Index)and NDVI(normalized difference vegetation index)indicators are calculated,and the mangrove forest mask range is extracted by applying the integrated NDWI-NDVI index threshold method.Mangrove forests are subsequently extracted by using four supervised classification methods:maximum likelihood,neural net,support vector machine and Mahalanobis distance.The classification accuracy and Kappa coefficient are obtained by calculating the confusion matrix.Simulation results show that the method proposed in this paper has a good sample separation effect,the construction sample points can well reflect the spe￣ctral information of mangrove forests and can meet the requirement of multiple supervised cla￣ssification methods,and high-precision mangrove forest remote sensing extraction is finally achieved.

关 键 词：红树林遥感 高光谱数据 归一化水体指数 归一化植被指数 掩膜提取 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]