基于植被分区的中国植被类型分类方法  

Methods of the National Vegetation Classification based on Vegetation Partition

作　　者：郝莹莹[1,2] 罗小波[1] 仲波[2] 杨爱霞[2] 

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]中国科学院遥感与数字地球研究所遥感科学国家重点实验室,北京100101

出　　处：《遥感技术与应用》2017年第2期315-323,共9页Remote Sensing Technology and Application

基　　金：国家863计划项目"星机地综合观测定量遥感融合处理与共性产品生产系统"(2013AA12A301);国防科工局高分观测专项课题"GF-4卫星图像辐射校正技术"(30-Y20A02-9003-15/16)

摘　　要：高精度的土地覆盖分类产品对定量遥感研究及遥感应用等具有非常重要的意义。目前免费的且全球覆盖的土地分类产品已有很多,但这些产品多为国外研究机构和人员所研发,由于对中国区域地形复杂、植被结构特征差异与农作物种植结构差异等没有进行充分的研究,使得这些产品用于中国区域的分类时其精度尤其是植被类型的分类精度较低。因此,生产一种针对中国区域的植被类型分类产品是非常必要的。针对中国区域地形、土壤等信息,并在借鉴现有的植被区划的基础上,发展了一种基于植被分区的中国植被类型分类方法,该分类方法以长时间序列为基础,能以较高的时间分辨率捕捉地表随时间变化的信息,从而利用地物在时间维上的差异提高分类精度,并利用该方法完成了2012年中国土地覆盖分类。此外还通过分层随机采样的方法对分类结果进行了精度评估,发现本分类产品的总体精度和Kappa系数有较大提高,其中本文产品总体精度为90.78%,Kappa系数为0.86;并通过与MODIS土地覆盖数据产品进行比较,发现该产品精度比MODIS土地覆盖数据产品在植被类型上提高了61.38%。The high precision land cover classification products has a very important significance for the study of quantitative remote sensing and remote sensing applications.Now there has a lot of free and global coverage land classification products,which is mostly developed by foreign research institutions and personnel.Due to the China regional complex terrain,differences in vegetation structures and in crop planting structures have not been fully studied,the classification accuracy of these products in the area of China is very low,especially the vegetation classification accuracy.Therefore,it is necessary to produce a vegetation type classification product of China region.So according to Chinese regional topography,soil and other information,and based on existing vegetation regionalization,we developed a method of Chinese vegetation classification based on vegetation partition.This method is based on long time series,which can capture the surface information changed with time with the high time resolution,and can improve the classification accuracy by the differences in the time dimension of the object.And we complete the 2012 national land cover classification by using the method.At last,the method of stratified random sampling was carried out to assess the accuracy of the classification results,found that the overall accuracy and Kappa coefficient of the classification products is greatly improved,which the overall accuracy was 90.78%,Kappa coefficient was0.86.And the product was compared with the MODIS land cover data products,found that the accuracy of this product increased by 61.38%in vegetation than the MODIS land cover data product.

关 键 词：土地覆盖分类 地形复杂 植被类型 植被分区 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]