Peanut planting area change monitoring from remote sensing images based on deep learning

As a powerful image processing technology, deep learning can extract representative and distinguishing features from remote sensing images in a hierarchical way. Considering the spectral information as the network input, the output spectral features from the network can be directly fed into a subsequent classifier to realize the classification based on pixel level. This paper presents a method based on convolutional neural networks (CNN) for peanut planting area extracting from Landsat-8 multispectral remote sensing images and applied in Zhenyang county in Henan province. The experimental results show that the architecture of CNN achieve good performance, the overall accuracy is 96.42%, kappa efficient is 0.944. Statistical calculations show that the peanut planting area of Zhengyang county increased by 12.1% in 2017 compared with 2015 which consistent with the current agricultural subsidy policy. The exploration of using CNN for crop recognition from multispectral images has excellent performance and great potential in agricultural remote sensing classification.