Extracting Winter Wheat Planting Area in Henan Province Based on Mixed Pixels Decomposition

The existence of mixed pixels in satellite images, especially those with coarse resolution, is still a challenge for crop planting area extraction with traditional classification methods. In this study, taking Henan Province as the pilot area, N_FINDR algorithm was used for automatic determination of spectral endmembers, and then Linear Mixing Model was utilized for mixed pixels decomposition to derive the planting area of winter wheat with multi-temporal MODIS data. Moreover, validation was carried out with statistical data. The results show that the phenological characteristic of winter wheat can be identified by the endmembers in general, which indicates that N-FINDR is an applicable algorithm for endmember determination. The extracted planting area of winter wheat in Henan Province is 4939.174 thousand hectares in 2008, with a total accuracy reaching up to 93.9%. The accuracies are various for different region; however, they can be improved with higher dimensions data and the mask data such as DEM or actual classification data.