Stoma Detection in Soybean Leaves and Rust Resistance Analysis.

Area of Science:

• Plant Science
• Agricultural Engineering
• Computational Biology

Background:

• Stomata are vital for plant immunity, and their morphology correlates with disease resistance.
• Accurate stomatal analysis is crucial for soybean breeding and disease resistance research.
• Traditional methods struggle with complex soybean leaf images, hindering stoma detection.

Purpose of the Study:

• To develop an advanced model for accurate soybean stoma detection and phenotypic parameter analysis.
• To enhance existing deep learning models for improved feature extraction of stomatal characteristics.
• To investigate the relationship between stomatal traits and soybean rust disease resistance.

Main Methods:

• Proposed a Soybean Stoma-YOLO (SS-YOLO) model integrating attention mechanisms (LSKA, DLKA) into YOLOv8 architecture.
• Modified Spatial Pyramid Pooling-Fast (SPPF) and Neck modules to capture multi-scale and irregular stomatal features.
• Applied the SS-YOLO model to analyze stomatal parameters (length, width, area, orientation, density, distribution) in soybean varieties under disease stress.

Main Results:

• The SS-YOLO model achieved a high detection accuracy of 98.7%.
• Effectively extracted and quantified stomatal features and related indices.
• Identified Dandou21 (DD21) as having stable stomatal morphology and distribution, correlating with rust disease resistance, unlike Fudou9 (FD9) and Huaixian5 (HX5).

Conclusions:

• The SS-YOLO model significantly improves soybean stoma detection accuracy and analysis.
• Stomatal characteristics are reliable indicators of soybean disease resistance.
• This approach offers a valuable tool for soybean breeding and plant disease resistance research.