Identification of Winter Wheat Plantation Based on Sentinel-2 Imagery

摘要

As an important food crop in the world, remote sensing identification of winter wheat (Trilicum Aestivum L.) and extraction ofplanting areas are the basis ofgrowth monitoring, disaster assessment and yield prediction. ESA Sentinel-2A/B satellite has significant advantages in crop identification with its high spatial resolution and short revisit period. In this study, northern and central Anhui, the main wheat production area in Anhui Province in China, were selected as the study area to explore the identification algorithm of winter wheat under different planting conditions. Based on the variation curve of the normalized difference vegetation index (NDVI) time series in the key phenological period of winter wheat, the specificity of its greenness was explored, and the screening rules were constructed accordingly. The spatial distribution of winter wheat in the study area was screened out by Decision Tree algorithm, and the extraction effect was improved by using a cultivated land distribution product. The extraction effect was evaluated by means of random samples, and the extraction area was tested with reference to the county-level wheat planting area data recorded in the statistical yearbook. The results showed that the estimation accuracy of winter wheat area in northern Anhui counties was 86.86%, and the Kappa coefficient was 0.80 (the Omission was 8.48%, the Commission was 9.58%). The estimation accuracy of winter wheat area in central Anhui counties was 80.00%, and the Kappa coefficient was0.62 (the Omission was 17.30%, the Commission was 40.20%). The overall extraction effect is inferior to that of the main producing areas in northern Anhui. This study provides a valuable reference for mapping wheat production under fragmented and complicated land surface conditions, which will further contribute to decision-making by government and relevant departments.