Extraction of Refined Crop Type Over Agriculture Region of Heilongjiang

The spatial distribution of fine crop types at regional scale is required by numerous research communities. The traditional approach with limited time-phases is hard to capture the signatures presented within different growth period of various crop types. With the improvement of understanding on the phenology feature of major crops and the accumulation of satellite-based observations, there is a chance to distinguish detailed crop clusters with elevated accuracy. In this work, we investigated the phenological feature of four dominant crops (soybean, wheat, maize, and paddy) in multi-spectrum space through ~800 representative crop samples within typical agriculture regions of Heilongjiang using MODIS daily surface reflectance product covering related growth period of 2005-2018. Features with the high degree of separation among land cover clusters are screened out to construct the model in identifying typical crop types in terms of weighted temporal features and classification scheme, which is applied to extract the crop map of Heilongjiang province in 2019. The results show higher accuracy achieved over main agriculture region of soybean, wheat, maize, and paddy, and reduced accuracy over field of wheat or other mixed crops at MODIS pixel scale. Our validation shows the overall accuracy of 0.9816 and kappa coefficient of 0.9702 through the comparison with ~3000 random selected ground sites. The preliminary application of the presented approach performs well via the capture on valid phenology features of major crops within dominant agriculture region of Heilongjiang, with the potential to serve the extraction of fine crop types over wide agriculture regions