data summary
This dataset is based on the Google Earth Engine cloud computing platform, using Sentinel-2 remote sensing data and random forest algorithm from 2017 to 2022 to obtain a dataset of spatiotemporal distribution of major crops (rice, corn, and soybean) at a resolution of 10m or 20m.
Data Sources
Google Earth Engine cloud computing platform, Sentinel-2 remote sensing images
Data generation or processing method
First, a multi-dimensional time series classification feature set was constructed based on the Google Earth Engine cloud computing platform and Sentinel-2 remote sensing images; a large number of ground sample points were collected by combining ground surveys and Collect Earth, Google HD images in historical periods and visual interpretation; then Using the random forest model and feature selection algorithm, the distribution map of crop types has been drawn year by year since 2017; finally, the classification model and classifier migration ideas in the historical period are used to realize the extraction of crop information in the year without samples.
data space projection
Projected Coordinate System：WGS_1984 _UTM_Zone_51N Geographic Coordinate System：WGS_1984
Data Quality Statement
Calculate and calculate the confusion matrix through the ground sample points to verify the data accuracy. The verification accuracy in 2017-2020 is: overall accuracy (OA) = 82.4%, 89.2%, 89.7%, 95.6%, Kappa coefficient = 0.775, 0.845, 0.852, 0.936 , 2021 due to the lack of ground sample points did not conduct accuracy verification. Through cross-validation by comparing with statistical data at the city level, the comparison results at the city level from 2017 to 2020 are: rice R2 is 0.98, 0.99, 0.99, 0.99, corn R2 is 0.91, 0.99, 0.98, 0.94 , soybean R2 were 0.96, 0.91, 0.96, 0.96.
Inquiry address: https://www.dilitanxianjia.com/