Sugarcane classification using the Chinese satellite image with high-resolution based on the PSENet algorithm

Sugarcane is a major sugar crop which support for the half world sugar provide with widely growth in Guangxi, Guangdong, Yunnan, and Hainan of China. Remote sensing technology provide an effective way which is a large area coverage in short periods for sugarcane classification which is developed from MODIS data (250m) to submeter satellite data to more precise result. Deep learning with a high performance was used in recent years which needs high-quality samples. The number of samples with wildly distribution is one of important factors for the training result which needs heavy workflow of labeling by manual work. In this research, a residual neural network, PSENet is proposed to high accuracy of sugarcane classification compared with the U-Net Model.