利用无人机高光谱影像的红树林群落物种分类  

Mangrove forest species classification based on the UAV hyperspectral images

作　　者：易俐娜[1] 张桂峰 魏征[4,5] 王冕卿 刘晋珂 王柳靖 YI Lina;ZHANG Guifeng;WEI Zheng;WANG Mianqing;LIU Jinke;WANG Liujing(School of Geoscience and Surveying Engineering,China University of Mining and Technology(Beijing),Beijing 100083,China;Aerospace Information Research Institute,Chinese Academy of Science,Beijing 100094,China;University of Chinese Academy of Sciences,Beijing 100190,China;South China Sea Institude of Planning and Environmental Research,Guangzhou 510145,China;Technology Innovation Center for South China Sea Remote Sensing,Surveying and Mapping Collaborative Application Ministry of Natural Resources,Guangzhou 510145,China)

机构地区：[1]中国矿业大学(北京)地球科学与测绘工程学院,北京100083 [2]中国科学院空天信息创新研究院,北京100094 [3]中国科学院大学,北京100190 [4]国家海洋局南海规划与环境研究院,广东广州510145 [5]南海遥感测绘协同应用技术创新中心,广东广州510145

出　　处：《测绘通报》2022年第11期26-31,共6页Bulletin of Surveying and Mapping

基　　金：高分遥感测绘应用示范系统(二期)(42-Y30B04-9001-19/21);国家自然科学基金(61405204);中央高校基本科研业务费项目-中国矿业大学(北京)(2022YQDC12);中国科学院战略性先导科技专项(A类)(XDA13020506);中国科学院科研仪器设备研制项目(YJKYYQ20170044);中国科学院重点研发项目(Y9F0600Z2F);国家重点研发计划(2018YFB0504903,2016YFB0501402);广东省促进经济发展专项资金粤自然资合[2020]012号。

摘　　要：近年来红树林群落中物种结构简单、功能退化等环境问题日趋严重,为了及时准确掌握红树林群落的物种空间格局与分布,本文首先基于深圳福田红树林自然保护区无人机高光谱影像,利用归一化差值植被指数和归一化潮间红树林指数提取植被区域;然后在植被区域根据最佳指数法选取信息量大、波段相关性小的波段组合,分别采用基于像素支持向量机分类方法和面向对象影像分类方法对红树林物种进行分类。试验结果表明,基于像素支持向量机分类方法的总体精度为81.03%;利用面向对象影像分类方法的总体精度为85.58%。面向对象影像分类方法能有效去除椒盐噪声,充分利用对象光谱、形状及纹理信息,提供更准确的红树林分布信息。In recent years, mangrove forest community species losses and functional degradation have become more and more serious. In order to timely and accurately extract the spatial pattern and distribution information of mangrove forest, this paper first extracts the vegetation area based on the UAV hyperspectral image of Futian mangrove nature reserve in Shenzhen using the normalized difference vegetation index and intertidal mangrove index, and then selected the band combination using the best index method. The pixel-based support vector machine classification(SVM) and object-oriented image classification(OOC) methods are used to accurately identify mangrove species. The experimental results show that the overall accuracy of SVM classification and OOC methods are 81.03%, and 85.58% respectively. In conclusion, The OOC methods can effectively remove the salt and pepper noise, makes full use of the spectral, shape and texture information of the object, and provides more accurate mangrove distribution information.

关 键 词：高光谱 无人机 红树林 支持向量机 面向对象 

分 类 号：P237[天文地球—摄影测量与遥感]