基于多种植被指数信息与联合稀疏表示的红树林种类识别  

Mangrove species classification based on multiple vegetation index extraction and joint sparse representation

作　　者：苏岫[1] 耿杰[2] 马晓瑞[2] 王洪玉[2] 王祥[1] SU Xiu GENG Jie MA Xiao-rui WANG Hong-yu WANG Xiangl(National Marine Environmental Monitoring Center, Dalian 116023, China School of Information and Communication Engineer- ing, Dalian University of Technology, Dalian 116024, China)

机构地区：[1]国家海洋环境监测中心,辽宁大连116023 [2]大连理工大学信息与通信工程学院,辽宁大连116024

出　　处：《海洋环境科学》2017年第1期114-120,共7页Marine Environmental Science

摘　　要：红树林种类识别对于研究红树林生态系统的变化具有重要意义。本文以广西铁山港红树林区为研究区域,以国产资源三号测绘卫星数据为数据源,分析区内各种红树林的光谱特性,并结合多项植被指数(RVI,NDVI,VARI和NDGI)信息,采用联合稀疏表示分类器进行红树林种类遥感识别。本文主要分析了桐花树、海漆、白骨壤、红海榄、秋茄、海桑、木榄这7种红树林种类以及陆地灌丛、泥滩、草地这些非红树林种类的几项植被指数,并结合多光谱图像的几何空间与光谱特征空间,采用联合稀疏表示算法进行红树林种类分类。利用组合光谱和4种植被指数信息进行分类可以达到最好的分类效果,总体精度为95.37%,kappa系数为0.9347。实验结果表明:光谱特征结合植被指数信息进行分类能提高分类精度,四种植被指数中NDVI对于区分红树林种类具有更大的贡献,联合稀疏表示分类器在红树林种类识别中表现出优异的分类效果。Mangrove species classification is important for studying the changes of mangrove ecosystem. In this paper, the distribution of mangroves in Tieshangang is chosen as the study area, and domestic ZY3 mapping satellite data is a- dopted as the data source. The spectral characterization of various mangroves is analyzed ,four different vegetation indi- ces （ RVI, NDVI, VARI and NDGI） are exacted to add vegetation information, and the joint sparse representation algo- rithm is utilized to distinguish seven mangrove species. We analyze the vegetation index of seven mangroves （ Ae- giceras corniculatum, Excoecaria agallocha, Avicennia marina, Rhizophora stylosa, Kandelia candel, Sonneratia apetala and Bruguiear gymnorrhiza） and other species （ bushwood, mudhank and grassland）. The geometric dimension and spectral dimension are combined and the joint sparse representation is used for classification. The overall accuracy reached 95.37% and kappa coefficient reached 0. 9347 when we use the spectral data incorporate with four vegetation indices. Experiments show that spectral features combined with vegetation indices can improve classification accuracy,and NDVI has a greater contribution than other indices to distinguish mangrove species. Furthermore,joint sparse rep- resentation classifier has a good performance in mangrove species classification.

关 键 词：红树林 稀疏表示 植被指数 资源三号 遥感分类 

分 类 号：X87[环境科学与工程—环境工程]