Answers are grounded in this paper and cite their sources.
Accurate identification of drought resistance is fundamental to drought-resistant germplasm screening and precision agricultural management. Existing methods remain insufficient in jointly modeling temporal evolution, natural growth differences, and stress-sensitive phenotypic responses under complex field conditions, which limits the analysis of multi-temporal UAV imagery. To address this issue, this study proposes a Multi-Dimensional Dynamic Difference Representation and Progressive Fusion Network (MDRF-Net) for UAV-based melon drought-resistance identification. MDRF-Net first uses a Temporal Collaborative Attention Fusion (TCAF) module to capture cumulative phenotypic responses across multiple observation dates, and a Dynamic Attention Convolution (DAC) module to enhance differential feature extraction between drought-stressed and well-watered states. A Multi-Scale Branch Fusion (MSBF) module and a Salient Feature Selection (SFS) module are further introduced to model multi-scale spatial patterns and perform stress-sensitive feature focusing. Experimental results show that MDRF-Net achieved an accuracy of 93.0% and an F1-score of 90.3%, outperforming representative machine learning and deep learning baselines. These results demonstrate the potential of MDRF-Net for high-throughput drought-resistant germplasm screening.
No takes yet. Share an insight, caveat, or question.
Gao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: