Classification Of The Cropland Using Landsat OLI Data By Decision Tree Method In Dancheng County,Henan Province

Remote sensing images can obtain crop planting information on the surface timely and accurately through digital processing,this is of great significance to how to make better use of the cultivated land ecosystem on which human beings depend.And it can provide important data sources for the analysis of crop planting types,crop yield estimation and optimization of cultivated land planting structure,is also important for national food security and food strategic reserves.The classification of crop types is a complex process and which needs to minimize errors and improve classification accuracy.So this paper selects Landsat OLI data as data source based on considerations of large observation extent,moderate time resolution and spatial resolution,and convenient data acquisition approach etc.This paper selects Dancheng County of Henan province(115°10?E-115°46?E,33°38? N-33°65? N)as the study area,selects the Landsat OLI data as data source,and determines crop Phenology through statistical yearbook data and field investigation data.The remote sensing image pre-processing tools in ENVI 5.2 software used to derive the multitemporal NDVI(Normalized Difference Vegetation Index band),NDBI(Normalized Difference Building Index)and NDWI(Normalized Difference Water Index)and decide the crop types classification threshold value for NDVI,NDBI and NDWI,respectively,through the repeated verification and modification works in IBM SPSS Modeler 18.0 software,then,decision tree method used to produce the final crop types classification maps.The accuracy of crop types classification results in the study area is evaluated based on confusion matrix,and the cause factor of crop types classification error is analyzed in study area.The results of the study show that the classification results of crop types based on the decision tree method are basically conform to the cultivated land structure in the study area.In some parts of the study area are characterized by the complexity of planting species,large number of plots,and with smaller area of the individual crops.Due to such limitation,the final crop types were classified with error of misclassification or omission.On the other hand,the image resolution also led errors,which were mainly caused by the mixed pixels.The overall accuracy and Kappa coefficient by confusion matrix of the accuracy evaluation are 64.65% and 58.25% respectively.To a certain extent,it can be considered that the technical route of this paper can meet the actual demand of fast and rough estimation of crop planting area.