Use 2022 Sentinel 2 optical satellite data as the data source.

Using the 2022 Sentinel 2 optical satellite data as the data source, the entire Xinjiang is divided into different areas and different types of layer layers according to cultivated land, desert, glacier, lake, forest land, grassland, impervious layer, bare land, and in accordance with the region and order. For each type of element, select samples within the entire Xinjiang, establish a small sample library, and obtain the extraction results of the entire Xinjiang element through sample enhancement and deep learning model selection. According to this method, deep learning and extraction are carried out on each ground class elements, and this ground class is used as a mask for the next ground class until all the required ground classes are completed. Finally, the results of information extraction are combined to obtain a complete LUCC classification map.

The data spatial resolution is high, the classification is rich in spatial details, and the categories are clear and complete. The overall accuracy OA of different elements is above 0.9, among which cultivated land, water bodies and glaciers have the highest overall accuracy, followed by grassland. The data meets the technical requirements.