Enhancing yield prediction in maize breeding using UAV-derived RGB imagery: a novel classification-integrated...

Area of Science:

• Agricultural Science
• Remote Sensing
• Data Science

Background:

• Accurate maize yield prediction is vital for food security and agricultural optimization.
• Unmanned Aerial Vehicle (UAV) derived RGB imagery offers a promising tool for crop monitoring.
• Integrating classification with regression can potentially improve yield estimation models.

Purpose of the Study:

• To develop and evaluate a novel classification-integrated regression approach for enhanced maize yield prediction.
• To compare the performance of different machine learning classifiers (SVM, DT, RF) for yield data categorization.
• To assess the effectiveness of combining yield classification with class-specific regression models.

Main Methods:

• UAV-derived RGB imagery was used to extract vegetation indices (VIs) across different growth stages.
• Support Vector Machine (SVM) was identified as the best classifier for categorizing yield into low, medium, and high classes.
• Two regression methodologies were compared: direct Random Forest (RF) regression and SVM classification followed by class-specific RF regression.

Main Results:

• The early vegetative growth phase showed the lowest prediction errors for yield estimation.
• The classification-integrated regression method (Method 2) significantly outperformed direct regression (Method 1).
• Method 2 reduced Root Mean Square Error (RMSE) by 45.1% in calibration and 3.3% in validation.

Conclusions:

• Combining classification (SVM) with regression (RF) enhances the precision of maize yield prediction using UAV data.
• This integrated framework provides a scalable tool for maize breeding programs and precision agriculture.
• UAV-based phenotyping holds significant potential for improving agricultural productivity and supporting global food systems.