Abstract
Rapid and non-destructive monitoring of maize SPAD is important for precision water and nitrogen management in arid irrigation areas. This study developed a UAV–smartphone cross-device calibration framework for SPAD estimation using UAV multispectral imagery, smartphone RGB images, and ground-measured SPAD data collected at the jointing, tasseling, and grain-filling stages. Spectral and texture features were integrated for UAV-based modeling, and regularized least-squares calibration was used to map smartphone-derived features into the UAV feature space. The random forest model using combined spectral and texture features achieved the best UAV-based performance, with validation R² values of 0.81, 0.79, and 0.75 across the three growth stages, representing an improvement of more than 9% over single-feature models. After cross-device calibration, smartphone-based SPAD estimation achieved R² values of 0.80, 0.86, and 0.82, respectively. These results demonstrate that cross-device feature calibration can effectively bridge UAV multispectral and smartphone RGB observations, providing a low-cost and accurate approach for field-scale maize SPAD monitoring and precision water–nitrogen management.