Remote Sensing Mapping Of Staple Crops Distribution In Jilin Province

The timely and accurate acquisition of crop spatial distribution and planting area is of great significance for the government to adjust agricultural structure and formulate agricultural policies and achieve sustainable agricultural management.Remote sensing technology has the ability to quickly detect crop distribution information on a macro scale,making it one of the main methods for current crop planting area detection.With the continuous development of remote sensing technology,the number of remote sensing image data sources is constantly increasing,and the image resolution is also continuously improving.The acquisition of large-scale,high-resolution crop spatial distribution information requires the download,storage,processing,and analysis of massive remote sensing images.The advent of cloud platforms provides an opportunity to quickly and accurately obtain large-scale high-resolution crop distribution information.Based on the Landsat-8 and Sentinel-2 optical images and Sentinel-1 radar images of the Google Earth Engine cloud platform,this paper selects different classification methods for crop classification experiments on remote sensing images in Jilin Province,and conducts crop mapping research from three aspects of optical remote sensing,radar remote sensing and multi-source remote sensing.Using Landsat8 and Sentinel-2 data as data sources,research was carried out on crop classification methods using optical remote sensing.Vegetation index,elevation,and slope information were selected as auxiliary data for crop classification identification.In the Landsat-8 classification experiment,the first seven 30-meter resolution bands were selected as the basic spectral data,and the three vegetation indices of NDVI,NDWI,and NDBI were calculated from the spectral data.Classification feature set.For multi-dimensional classification feature sets,random forest,CART,and minimum distance classifiers are used for classification experiments.The results show that random forest classifiers are better than CART and minimum distance classifiers for crop recognition,and the classification accuracy reaches 93.88%.In the Sentinel-2 classification experiment,four 10-meter-resolution bands of red,green,blue,and near-infrared were selected as the basic spectral data,and the two vegetation indices of NDVI and NDWI were calculated from the spectral data.The vegetation index,elevation,and slope information and spectrum were Combine the data to build a multi-dimensional classification feature set.For the multi-dimensional classification feature set,random forest,CART,and minimum distance classifiers are used for classification experiments.The results show that random forest classifiers are better than CART and minimum distance classifiers for crop recognition,and the classification accuracy reaches 93.89%.Compared with the other two classification methods,the random forest algorithm is more suitable for crop remote sensing mapping.Using Sentinel-1 data as a data source,research on crop classification methods using radar remote sensing was carried out.Select 30 days as the time interval for image synthesis to ensure that monthly radar image data is used for crop classification experiments throughout the entire growing period(May-October)of the crop.The Sentinel-1 radar images in different months were classified respectively.After comparing the accuracy of the generated classification results,it was found that the corn,soybean and rice in the Sentinel-1 radar images in August had good separability,and the overall classification accuracy reached 84.82.%.In experiments that explored the identification of crops with different polarization methods,it was found that compared with cross-polarization,the same-polarity image has better recognition of crops.The random forest classifier is selected to classify multiple-time same-polarity radar images with the highest overall classification accuracy.It is 88.72%.Multitemporal bipolar data Sentinel-1 data were used to perform random classification experiments using random forest,CART,and minimum distance classifiers.The results show that random forest classifiers are better than CART and minimum distance classifiers for crop recognition,and the classification accuracy has reached 90.48%.Using Sentinel-1 and Sentinel-2 data as data sources,research on multi-source remote sensing crop classification methods was carried out.Random forest,CART,and minimum distance classifiers were used for classification experiments.The results show that the random forest classifier is superior to CART and minimum distance classifiers in crop recognition,and the classification accuracy reaches 95.33%.Based on the GEE cloud platform,Landsat-8,Sentinel-1,Sentinel-2,and multi-source images were selected to identify crops in Jilin Province,and the classification accuracy was 93.88%,90.48%,93.89%,and 95.33%,which indicates that the GEE cloud platform can be used.Realize high-precision and high-efficiency extraction of the distribution information of bulk crops in Jilin Province.