Land Use/cover Classification Based On Landsat/MODIS Data Fusion

At present,land use/cover change information has become an indispensable and important basic data for global change and sustainable development research.This is due to the fact that surface coverage is a combination of natural and artificial structures that can be observed on the surface of the land and is the result of the joint action of nature and human activities.With the rapid changes in global land use/coverage and the rapid development of remote sensing information technology,accurate and effective acquisition of high spatial-temporal resolution remote sensing images plays a crucial role in monitoring land use/cover information.In a large range of land cover classification studies,it is difficult to meet the requirements of high spatial-temporal resolution for remote sensing images obtained by relying solely on certain sensor data.Therefore,how to effectively improve the classification accuracy of land-use cover types is still a hot topic at home and abroad.This paper selects Qinhuangdao as a research area,refers to the USGS,IGBP,FAO,and the land use classification system of the Chinese Academy of Sciences,and combines the natural and human characteristics of Qinhuangdao to establish a land use classification system that meets the actual conditions in the region.Combining the advantages of MODIS high time resolution and high spatial resolution of Landsat OLI,based on the ESTARFM spatio-temporal data fusion model,high-temporal resolution NDVI data of 16 days and 30meters in Qinhuangdao in 2015 is generated.Using the dual logistic function fitting algorithm in TIMESAT3 software,the phenological parameters of the study area are extracted,which can better reflect the phenological characteristics of vegetation such as cultivated land,forest land,shrubs,and grassland.Obtain DEM data and derived slope,aspect data,NDBI(normalized difference building index),MNDWI(improved normalized difference water index),and NDSI(soil brightness index),and build a multi-feature classification data set,according to the separability analysis,the best combination of classification features is determined,and then the classification results are obtained using the object-oriented support vector machine classification method,and the classification accuracy is verified by the confusion matrix.The results show that:(1)The ESTARFM_NDVI data obtained by using the ESTARFM spatial-temporal fusion model has a high correlation with real Landsat_NDVI data,R~2 can reach 0.88,and the prediction results can be used for land use cover classification research.(2)The phenological parameters extracted by TIMESAT software can effectively reflect the phenological change characteristics of vegetation in the study area,thus greatly improving the classification accuracy of vegetation.(3)Using the method of object-oriented support vector machines to extract land use/cover information from the study area,the overall classification accuracy can reach87.15%,and the Kappa coefficient is 84.29%,among which the classification accuracy of woodland,shrub and grassland is 86.38%,81.59%and 83.51%respectively,it shows that the multi-feature parameters can be used to achieve high precision,and can significantly improve the accuracy of forest,irrigation and grass automatic classification.The innovation points are as follows:(1)Based on the ESTARFM spatial-temporal data fusion model,this paper uses MODIS data to supplement the 2015 missing Landsat data,and constructs a Qinhuangdao area high-r-e solution image(16day,30m)NDVI dataset.(2)In this paper,the vegetation phenology parameters and other classification characteri-stic parameters of Qinhuangdao area high spatial resolution are extracted,and the phenologic-al changes of the surface cover in the study area are analyzed,and the optimal classification f-eature combination is selected.The object-oriented support vector machine(SVM)method is used to extract Qinhuangdao area land use cover classification information.