Monitoring the growth status of rice based on hyperspectral satellite remote sensing data

This study proposes a novel approach to rice growth monitoring using a 3D Convolutional Neural Network (3D-CNN) model applied to hyperspectral satellite remote sensing data. The model combines spatial, temporal, and spectral information processing to enhance the accuracy of rice growth monitoring over large areas. A new loss function is introduced to address imbalanced yield label distribution. The model's performance is validated using rice yield data from China's main rice-growing regions, demonstrating superior predictive capability compared to existing methods. This approach offers a promising tool for improving food security through more accurate and timely crop monitoring.