Sat-Crop transformer: integrating temporal satellite imagery and GAN-based synthesis for vegetation health monitoring

Abstract

Monitoring vegetation health is vital for environmental management, agriculture, and land use planning. We introduce an innovative approach that integrates satellite imagery analysis and predictive modeling to predict vegetation indices and generate synthetic field images. We propose the Sat-Crop Transformer, a transformer-based model specifically designed for satellite crop monitoring for the prediction of crop health. Historical satellite imagery data, including vegetation indices such as the modified soil-adjusted vegetation index and environmental parameters, are used to predict future vegetation indices. The Sat-Crop Transformer demonstrates better performance in capturing long-range dependencies in temporal satellite data. A generative adversarial network–based approach is used to generate synthetic field images corresponding to the predicted vegetation indices, thereby enhancing the data visualization and model training. This integration of advanced predictive modeling and image generation provides a deeper understanding of environmental processes and supports decision-making in various domains.

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.