Monitoring maize phenology using multi-source data by integrating convolutional neural networks and Transformers

Abstract：Maize is an essential grain crop in China, playing a crucial role in safeguarding in national food security. However, the increasing instability of the maize cultivation environment caused by global climate change, along with various adverse stress factors, presents significant challenges to maintaining yield stability. Effective monitoring of maize phenology under stress conditions is crucial for optimizing agricultural management and mitigating yield losses. This study proposes an innovative phenological monitoring model utilizing near-ground remote sensing technology. High-resolution imagery of maize fields was collected using unmanned aerial vehicles (UAVs) equipped with multispectral and thermal infrared cameras. By integrating these datasets with Convolutional Neural Network (CNN) and Transformer, the study developed a robust and efficient model that fuses multispectral, thermal infrared, and accumulated temperature datasets. The proposed model enables accurate inversion and quantitative analysis of maize phenological traits, offering critical insights to support agricultural management strategies and enhance crop yield stability under stress conditions. The results showed that the integration of multispectral imagery and accumulated temperature achieved an accuracy of 92.9%, while the inclusion of thermal infrared imagery further improved the accuracy to 97.5%. Additionally, UAV-based remote sensing offers superior spatial resolution and operational efficiency compared to manual observation methods in precision and scalability. This study highlights the potential of UAV-based remote sensing, combined with CNN and Transformer as a transformative approach for precision agriculture. It provides a valuable framework for advancing agricultural informatization and enhancing crop management.

Key words: Maize; Crop phenology; Deep learning; UAV；Multi-source data