Early Crop Mapping Based on Sentinel-2 Time-Series Data and the Random Forest Algorithm

Early-season crop mapping and information extraction is essential for growth monitoring yield prediction, it facilitates agricultural management rapid response to disasters. However, training classifiers by remote sensing classification features early prediction can be challenging, as early-season only use image data during part of the period. In order overcome this limitation, study takes Sanjiang Plain an example investigate earliest identification time rice, maize soybean based on Sentinel-2 time-series random forest algorithm. Crop was then performed. Following analysis importance approach subsequent normalization, optimal greater than or equal 0.5 have yielded quite results in mapping, their overall accuracy highest mapping. The observed improve 5% 10 20 days delay. addition, maize, were mapped at irrigation transplanting period (10 May), jointing stage (9 July) flowering (29 July), with 90.4%, 90.0% 90.9%, respectively. This shows that suitable selected well ability extract acreage within reproductive