A statistical method for high-throughput emergence rate calculation for soybean breeding plots based on field...

Area of Science:

• Agricultural Science
• Plant Breeding
• Remote Sensing

Background:

• Accurate soybean emergence rate statistics are crucial for smart breeding, but current methods struggle with low throughput, efficiency, and precision, especially in dense planting environments.
• Environmental factors and seedling overlap complicate traditional counting methods, necessitating innovative solutions for high-throughput phenotyping.

Purpose of the Study:

• To develop and validate an effective and precise statistical method for counting soybean seedlings using Unmanned Aerial Vehicle (UAV)-scale and ground measurement data.
• To enhance the throughput, efficiency, and accuracy of soybean breeding screening under intensive planting conditions.

Main Methods:

• A combined background segmentation method using contrast enhancement filtering, ultra-green eigenvalues, and the Otsu algorithm was employed for image preprocessing.
• Deep learning object detection models (specifically Yolov8n) were utilized for soybean seedling identification and labeling.
• Novel algorithms, including 'growth normalization' and an 'inter-seedling occlusion counting algorithm' with a soft strategy for bounding box overlap, were developed for accurate seedling quantification.

Main Results:

• The proposed method achieved an overall accuracy of 99.18% in counting soybean seedlings, with an error rate of 0.82%.
• Yolov8n demonstrated strong performance with a mean Average Precision (mAP) of 85.15% for seedling detection.
• The background segmentation technique improved the mAP by 4.06%, highlighting its effectiveness in enhancing detection accuracy.

Conclusions:

• The developed UAV-based method provides robust support for statistical analysis of soybean emergence rates in intensive planting scenarios, significantly accelerating the breeding process.
• This innovative approach offers new insights and a valuable reference for developing efficient screening techniques in plant breeding.