UAV-based multi-sensor data fusion and machine learning algorithm for yield prediction in wheat.

Area of Science:

• Agricultural Science
• Remote Sensing
• Machine Learning

Background:

• Accurate early prediction of wheat grain yield is crucial for effective breeding programs.
• Unmanned aerial vehicle (UAV)-based multi-sensor data offers a high-throughput phenotyping approach.
• Integrating diverse sensor data with machine learning can enhance yield prediction accuracy.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms for wheat grain yield prediction using fused multi-sensor UAV data.
• To compare the predictive performance of individual sensors versus data fusion across different ML models.
• To assess the benefits of ensemble learning in improving yield prediction accuracy.

Main Methods:

• Utilized five machine learning algorithms: Cubist, Support Vector Machine (SVM), Deep Neural Network (DNN), Ridge Regression (RR), and Random Forest (RF).
• Employed data fusion techniques combining RGB, multi-spectral, and thermal infrared data from a low-cost UAV platform.
• Evaluated prediction accuracy using coefficient of determination (R²), Root Mean Square Error (RMSE), Residual Prediction Deviation (RPD), and Ratio of Prediction Performance to Inter-quartile Range (RPIQ).

Main Results:

• Multi-sensor data fusion significantly improved yield prediction accuracy compared to individual sensors across all tested ML models.
• Ensemble learning, integrating multiple ML models, achieved the highest prediction accuracy with an R² value up to 0.692.
• The best model achieved an RMSE of 0.916 t ha⁻¹, RPD of 1.771, and RPIQ of 2.602, indicating robust predictive capability.

Conclusions:

• Low-altitude UAV-based multi-sensor data, combined with data fusion and ensemble learning, provides a highly accurate method for early wheat grain yield prediction.
• This high-throughput phenotyping approach is valuable for enhancing the efficiency of selection in large-scale wheat breeding activities.