Classification and Verification of Land Use/Cover in Australia Using MODIS Time-series Data
基于MODIS时序数据的澳大利亚土地利用/覆被分类与验证

his study is to classify land use/cover of Australia in 2000 based on data MOD13Q1,for it's sensitive to climate change.In order to reconstruct high quality NDVI time series data,Savizky-golay filtering method was applied for the NDVI data set.In this study,a combination of unsupervised classification of ISODATA and hierarchical decision tree was applied for the land use/cover classification over the whole country of Australia,and NDVI values of specific season,transformed NDVI values,and principal component analysis were also performed on the data set for decision tree threshold value choosing.Based on area of contract and spatial location matching to assess the accuracy of the classification,comparison the classification results of MOD12Q1,GLC_2000 with the results was done for this study.Random selected samples and generate confusion matrix.The overall classification accuracy of 63.65% and the Kappa coefficient of 0.56 were achieved in this study.This result is better than the result achieved in MOD12Q1 and GLC_2000.