Deep Multiview Image Fusion for Soybean Yield Estimation in Breeding Applications

Area of Science:

• Agricultural Science
• Computer Science
• Genetics

Background:

• Accurate seed yield estimation is crucial for soybean cultivar development.
• Traditional methods are time-consuming and labor-intensive.

Purpose of the Study:

• To develop a machine learning (ML) approach for soybean pod counting.
• To enable genotype seed yield rank prediction using in-field video data.

Main Methods:

• A multiview image-based yield estimation framework utilizing deep learning architectures.
• Fusion of plant images captured from multiple angles.
• Comparison with manual pod counting and yield estimation.

Main Results:

• The ML framework accurately estimates soybean yield and ranks genotypes.
• Demonstrated efficacy in both controlled and field test plot environments.
• Significant reduction in time and human effort compared to manual methods.

Conclusions:

• Machine learning models show great promise for soybean breeding decisions.
• The developed framework opens new avenues for cultivar development.
• Accelerated breeding decisions through automated yield estimation.