Study On Crop Classification Of Guangxi Zhuang Autonomous Region Based On MODIS Time Series Data

China's subtropical region is mainly located in the southern of China,Guangxi Zhuang Autonomous Region because of its geographical location and terrain features can be used as a typical representative of China's subtropical mountain,in view of this paper,Guangxi as a research area that can better study the subtropical regions of China crops Classification.Guangxi Zhuang Autonomous Region is an agricultural province,because of the unique natural environment conditions of its wide range of agricultural crops,agricultural crops for rapid,accurate and efficient identification and classification of agricultural development and economic development of the inevitable demand,and a region of food security,Socio-economic,ecological functions and the corresponding policy development will have an important impact.In this paper,the Guangxi Zhuang Autonomous Region is a typical representative of the subtropical mountainous region of China.The Guangxi Zhuang Autonomous Region is the research area.Some of the crops and some special crops in the Guangxi Zhuang Autonomous Region are studied in this paper.MODIS 250 m from January 2013 to December 2013 Resolution MOD13Q1 EVI(enhanced vegetation index)and NDVI(normalized vegetation index)as data sources,field research and GoogleEarth expansion method for crop sample data collection.Firstly,the Timesat software is used to compare the denoising degree of the AG(Asymmetric Gaussian filter)algorithm and the SG algorithm to the denoising degree of the research area and the research object.Then,the algorithm of the study of the regional phenological index extraction is selected.Secondly,the SG filter algorithm is used to analyze the EVI time series And the NDVI time series were reconstructed and the phenological phenological indexes were extracted.Eleven phenological indicators(Start of season,End of season,Length of season,Base value,Position of middle of season,Maximum of fitted data,Amplitude,Left derivative,Right derivative,Large integral,Small integral).Finally,the sample data are divided into training sample data and test sample data.The random forest classification algorithm in EnMAP-Box is used to reconstruct the image of EVI phenology and reconstructed images,NDVI phenological index images and reconstructed images and other Crop classification and precision evaluation with different combinations.The study concluded that:(1)Based on the extraction of the original EVI and the NDVI index of the crop in 2013 and the comparison of the curves,it was found that different crops showed different vegetation index fluctuation characteristics,and different growth curves could be extracted according to different crops to determine the vegetation using MODIS Index the feasibility of categorizing crops.(2)Using Timesat to reconstruct the vegetation index time series and the extraction of phenological indicators,it is found that the S-G filter algorithm is more suitable for the extraction of crop phenology indexes in Guangxi Zhuang Autonomous Region than the AG algorithm and DL algorithm,And the NDVI time series were smoothed.The results showed that the growth curves of different crop fitting and reconstruction were different and could be used as the characteristics of crop classification.(3)Using the EnMAP-Box software to set the corresponding parameters to establish EVI RFC model and NDVI RFC model,from the applicability analysis,EVI RFC model F1 accuracy is higher than the NDVI RFC model F1 accuracy.The EVI RFC model is more moderated than the performance of the NDVI RFC model from the differences between the F1 accuracy curves of different crops.Most of the crops in the EVI RFC model began to stabilize after 11 classification trees,and the NDVI RFC model began to stabilize around the 20 classification trees.The applicability of the EVI RFC model to the crop in the study area is better than the NDVI RFC model.From the analysis of the importance of variables,the main variables of the EVI RFC model are the growth season small integral variables,the growth base value variables,and the growth season starting variables.The main variables of the NDVI RFC model are the medium-term variable in the middle-growth season,the maximum variable of the fitting function,and the base value of the growth season.(4)The relationship between the overall classification accuracy of vario us images and the combined image and the kappa precision was obtained by usin g EVI,NDVI vegetation index phenological index image,fitting reconstructed im age and combination of the two,and the precision evaluation:EVIph+EVIfit+NDV Iph+NDVIfit>EVIfit+NDVIfit>NDVIIF>NDVIfit>NDVIfit>EVIph+EVIfit>EVIIF The resu lts show that the combination of the phenological index image and the reconst ructed image is more accurate than using one of the classification results al one.