基于MODIS EVI序列典型指标组合的土地覆盖分类研究  

Land Cover Classification Based on Typical Indices Combinations of MODIS EVI Time Series

作　　者：杜子涛[1,2] 杨小明[1,2] 颜树强[1,2] 任海峰[1,2] 王德军[1,2] 

机构地区：[1]河北工业大学土木工程学院,天津300401 [2]河北省土木工程技术研究中心,天津300401

出　　处：《海洋测绘》2011年第6期20-23,共4页Hydrographic Surveying and Charting

基　　金：国家重点基础研究发展计划项目(2007CB714406);国家科技支撑计划(2008BAC34B03);欧盟项目CEOP-AEGIS(FP7-ENV-2007-1Grantnr.212921)

摘　　要：以2008年经Savitzky-Golay滤波技术平滑处理的MODIS EVI时间序列数据为数据源,采用支持向量机分类法对年EVI矢量模、年EVI最大值和年EVI最小值3个指标组成的影像进行土地覆盖分类,并利用验证样本对分类结果进行精度评价。试验结果表明:采用选用的3个指标不仅能减少分类的数据量、加快分类速度,而且还能减少其他无效信息对分类的干扰,提高分类的精度。By using MODIS EVI （16-day,250m） time series smoothed by Savitzky-Golay filter in 2008, this work chose modulus of vector, maximum and minimum EVI three indices in one year to do classification. Then the land cover map was generated using support vector machine （SVM） classification method. The classification accuracy was assessed using validation samples and the result indicates that three classification indices used in the paper could not only reduce the data volume and quicken the speed of classification, but also could reduce the disturbance of other invalidation information to classification and get better classification accuracy.

关 键 词：影像分类 EVI时间序列 矢量模 最大值 最小值 

分 类 号：TP7[自动化与计算机技术—检测技术与自动化装置]