植被遥感分类方法研究进展  

Research Progress on Remote Sensing Classification of Vegetation

作　　者：杨超 邬国锋[1,3] 李清泉[1] 王金亮 渠立权[5] 丁凯[1,2] YANG Chao;WU Guo-feng;LI Qing-quan;WANG Jin-liang;QU Li-quan;DING Kai(Key Laboratory for Geo-Environmental Monitoring of Coastal Zone of the National Administration of Surveying, Mapping and GeoInformation&Shenzhen Key Laboratory of Spatial Smart Sensing and Services,Shenzhen University,Shenzhen 518060;College of Information Engineering,Shenzhen University,Shenzhen 518060;College of Life Sciences and Oceanography,Shenzhen University,Shenzhen 518060;College of Tourism and Geographical Sciences,Yunnan Normal University,Kunming 650500;School of Geography and Survey&Urban and Rural Planning,Jiangsu Normal University,Xuzhou 221116,China)

机构地区：[1]深圳大学海岸带地理环境监测国家测绘地理信息局重点实验室&空间信息智能感知与服务深圳市重点实验室,广东深圳518060 [2]深圳大学信息工程学院,广东深圳518060 [3]深圳大学生命与海洋科学学院,广东深圳518060 [4]云南师范大学旅游与地理科学学院,云南昆明650500 [5]江苏师范大学地理测绘与城乡规划学院,江苏徐州221116

出　　处：《地理与地理信息科学》2018年第4期24-32,共9页Geography and Geo-Information Science

基　　金：国家自然科学基金重大研究计划培育项目"基于多源时空大数据群体时空移动规律挖掘与动力学建模"(91546106);国家重点研发计划"典型脆弱生态修复与保护研究"重点专项(2017YFC0506200;2017YFC0506201);深圳市经信委"创新链+产业链"融合专项扶持计划项目(201507211219247860)

摘　　要：遥感影像分类是获取植被信息的一个重要方式,而分类方法的选择是影响分类精度的关键因素。从植被遥感分类中是否具有先验知识的分类方法、结合多源遥感数据及相关辅助信息的分类方法、基于机器学习的分类方法和其他分类方法等方面对国内外相关研究进行了评述。为深入研究在现有分类方法基础上如何提高植被遥感分类的精度及进行植被遥感细分类,建议利用影像融合DEM的三维地形辅助分类及选取训练样本提高样本分离度;将混合像元分解思想融入到现有的植被遥感分类方法中,综合地形、纹理、光谱等辅助信息进行细分类;继续开展对"软"分类器方法的研究,将单个像元分解为不同组分,从"亚像元"级进行植被遥感分类。Remote sensing classification is an important part of getting vegetation information,and the classification method is the most important factor to affect the classification accuracy.The research on vegetation remote sensing classification at home and abroad is reviewed from the aspects of traditional methods(supervised classification and unsupervised classification),the special methods(the classification method based on vegetation index,classification method using multi-temporal information,classification method integrating multi-source remote sensing data,texture,terrain,spectrum or other auxiliary information),the machine learning methods(intelligent classification,SVM,decision tree),the advanced methods(object-oriented classification,"soft"classification).To further research how to improve the accuracy of remote sensing classification for vegetation,it is worth recommending to use 3D terrain aided classification and select training samples from different perspectives which improve training sample separability;put mixed pixel decomposition method into the existing vegetation remote sensing classification,and integrate terrain,texture,spectrum and other auxiliary information to conduct vegetation subdivision;continue to research the"soft"classifier method,break through the traditional based pixel classification,unmixing a single pixel into different components.

关 键 词：植被分类 遥感 三维地形 混合像元分解 研究进展 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]