A Fine Classification Mapping Method for Crops Based on Time Series Chinese Gaofen Analysis Ready Data

Fine classification of crops is of great significance for agricultural planning, and sustainable development. This work uses Chinese GaoFen (GF) analysis ready data (ARD) products with high spatiotemporal resolution, taking a part of Huantai, Shandong Province as the research area. By constructing a prior knowledge base of time-space-spectrum, initial samples are extracted, and the number of samples is further increased through space-time expansion methods. High precision crop recognition results are obtained through training with a random forest model. In addition, the distribution of farmland blocks was overlaid to obtain a more refined crop classification map. The overall accuracy is proved up to 90%. This paper provides a high-precision, efficient, and more refined framework for crop classification mapping, which can be extended to achieve automated crop classification mapping with a larger range and longer time series.