arXiv subjects

Shafqaat Ahmad

Publications and source records attributed to Shafqaat Ahmad.

2 recordsLinked to original sources

arXiv2026

Embedding NDRE Trajectories into Contrastive Learning for Label-Free, Physiology-Aware Crop-Stress Staging and DSS Outputs

Timely detection of crop stress is critical for sustaining yields under increasing drought frequency, yet conventional vegetation index thresholds or image-based clustering often fail to capture stress progression, limiting their value for farm decision-making. To address this gap, we present EigenCL, a physiology-guided contrastive learning framework that stages crop stress from Sentinel-2 NDRE trajectories, with the goal of providing interpretable and transferable stress diagnostics for decision support systems (DSS). EigenCL was trained on 10,000 maize NDRE patches from drought-affected Iowa fields in 2020 and tested on Nebraska fields in 2023 without retraining, with validation incorporating soil-moisture records, U.S. Drought Monitor maps, and county-level yield statistics. The model produced four physiologically coherent stress clusters (Healthy, Mild, Moderate, Severe), significantly outperforming baselines including K-Means, SimCLR, ProtoCLR, and an ablation model (Silhouette = 0.748, DBI = 0.35, CHI = 49,624). Clusters aligned with maize growth stages, with severe stress peaking around tasseling-silking (VT-R1), a stage known to drive yield loss; moreover, EigenCL clusters correlated with soil moisture at 0-14-day lags (rho up to 0.72) and matched yield anomalies in drought-affected counties. By embedding NDRE trajectory dynamics into contrastive learning, EigenCL enables early stress alerts and interpretable DSS outputs (e.g., heatmaps, scouting priorities, regional risk indices), extending beyond single-date NDRE thresholds and supporting scalable monitoring for climate-smart agronomy.

cs.CV↗

arXiv2025

Temporal Vegetation Index-Based Unsupervised Crop Stress Detection via Eigenvector-Guided Contrastive Learning

Early detection of crop stress is vital for minimizing yield loss and enabling timely intervention in precision agriculture. Traditional approaches using NDRE often detect stress only after visible symptoms appear or require labeled datasets, limiting scalability. This study introduces EigenCL, a novel unsupervised contrastive learning framework guided by temporal NDRE dynamics and biologically grounded eigen decomposition. Using over 10,000 Sentinel-2 NDRE image patches from drought-affected Iowa cornfields, we constructed five-point NDRE time series per patch and derived an RBF similarity matrix. The principal eigenvector explaining 76% of the variance and strongly correlated (r = 0.95) with raw NDRE values was used to define stress-aware similarity for contrastive embedding learning. Unlike existing methods that rely on visual augmentations, EigenCL pulls embeddings together based on biologically similar stress trajectories and pushes apart divergent ones. The learned embeddings formed physiologically meaningful clusters, achieving superior clustering metrics (Silhouette: 0.748, DBI: 0.35) and enabling 76% early stress detection up to 12 days before conventional NDRE thresholds. Downstream classification yielded 95% k-NN and 91% logistic regression accuracy. Validation on an independent 2023 Nebraska dataset confirmed generalizability without retraining. EigenCL offers a label-free, scalable approach for early stress detection that aligns with underlying plant physiology and is suitable for real-world deployment in data-scarce agricultural environments.

cs.CV↗