Land Use Classification Based On The Time Series MODIS Data Taking Hedong Region Of Gansu Province As An Example

It is a fundamental project in remote sensing technology for applying remote sensing to classify land use.Remote sensing data has some advantages such as comprehensiveness and timeliness so that it can acquire rapidly and accurately the information of land use.MODIS data has a high spectral resolution and a clear spatial resolution.Therefore,it is a strong superiority and extensive application value for the classification of land use at macro level.In study region of Hedong area in Gansu province,this paper makes use of CART classification method based on MODIS data in 2016.It exploits potentialities for classification of land use in using MODIS data.Research on classification of land use in study region,main conclusions are as followed.Firstly,it plays a crucial role on extracting classification parameter features for smoothing carves of NDVI time series.The paper deals with the fitting of NDVI time series by means of Savitzky-Golay smoothing in TIMESAT software.It can keep original carve shape and clearly reflect growth condition of vegetation in a whole year.Secondly,selecting classification parameter features has a direct influence on the result of classification accuracy.There are six parameter features such as the NDVI growth rate in 18 to 28 phases,the time span of higher vegetation cover,and the NDVI mean value in 18 to 30 phases,the NDVI minimum value in whole year,the NDVI maximum value in whole year and NDVI range in whole year.In addition,the paper applies the classification method of CART in land use and land cover classification based on ENVI software.The threshold value is determined in extracting six parameter features and establishes 20 rules of CART on research classification of land use.Finally,it establishes seven land categories of land use based on MODIS data with time series.The classification accuracy of land use applies training sample ofMODIS data through checking based on the data of Landsat8 OLI.The total precision of classification is 82.47%,and the Kappa coefficient is 0.8018,Among the precision of forest land is 85.49%;the precision of garden land is 82.74%;the precision of arable land is 73.83%;the precision of grassland is 77.6%;the precision of water is89.92%;the precision of construction land is 89.73%;the precision of other field is94.1%.The classification result clearly reflects the actual condition in land use classification.Selecting multi-temporal data of MOD09Q1,it obtains a better classification result by the two bands reflectance data of MODIS.This method has an advantage of land use classification using multi-temporal data of MODIS data.The result shows that MODIS data has practical value and application capability on land use,and it has significance of the classification of land use in a large area.