Cereal Crop Monitoring Using Sentinel Time Series Through Google Earth Engine

In this study, we propose an approach based on Google Earth Engine (GEE) divided into two distinct steps. First, we use vegetation indices such as Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Leaf Area Index (LAI) and Harvest Index (HI). NDVI was used to characterize the growth stages of cereal crops and Otsu thresholding to extract cereal crops. The second step involved associating multitemporal datasets, containing a combination of spectral bands and vegetation indices, as input data for random forest (RF) machine learning classifiers to identify cereal crops. The classification accuracies reveal the show that the RF-NDVI provides an overall of 86.6% for cereal crop mapping, outperforming other techniques employed in this study, notably NDVI thresholding (70.7%), RF-EVI (80%), and RF-LAI (80%).