A dataset of land use types in the Lijiang River Basin from 2016 to 2020

This data set takes sentinel-2 of ESA from 2016 to 2020 as the original data source, and there are a total of 5 periods (2016, 2017, 2018, 2019, 2020) of image data of 10 scenes. First, we need to perform radiometric calibration and atmospheric correction on the original image. These processes are mainly completed in the plug-in sencor-20.09.00. After processing, we get an image after radiometric calibration and atmospheric correction. Then, in the snap software of ESA, the image is resampled and format converted, the image format is set to the image format suitable for ENVI opening, and the resolution of the band is adjusted to 10m. After processing, a preprocessed sentinel-2 satellite image is obtained. Since the Lijiang River Basin occupies two images, it is necessary to mosaic the two images in envi software, then cut the images according to the vector range of the Lijiang River Basin, and then fuse the green, blue and near-infrared bands into false color images. Based on the samples of the image interpretation, the training samples required for supervised classification are selected. After selecting the training samples, the maximum likelihood method is used for supervised classification, and the preliminary classification result map is obtained. Because the preliminary classification result map has a certain "pepper and salt" phenomenon, it needs to be optimized in the later stage, that is, manually correct the pixels with large deviation, and finally use the cluster analysis method to remove the error pixels in a small range to get the final classification result map. The classification result data set is saved in a compressed package in TIF format, with names in the order of year and land_ Use as the file prefix.