Automated stomata detection in oil palm with convolutional neural network

Area of Science:

• Agricultural Science
• Plant Physiology
• Biotechnology

Background:

• Stomatal density is crucial for identifying drought-tolerant oil palms.
• Current methods for measuring stomatal density are time-consuming and labor-intensive.
• Automating this process is essential for efficient breeding programs.

Purpose of the Study:

• To develop an automated system for accurate stomatal density measurement in oil palms.
• To create a specialized stomata detection model for oil palm leaf samples.
• To accelerate the selection of drought-tolerant oil palm genotypes.

Main Methods:

• Collected leaf samples from 128 oil palms across different age groups (nursery, juvenile, mature).
• Developed a convolutional neural network model using transfer learning on tiled micrographs.
• Tested the model on independent oil palm populations and cross-applied it to images from different conditions.

Main Results:

• The automated system achieved high detection accuracy across different oil palm populations (precision: 98.00-99.72%, recall: 96.88-99.62%).
• The model demonstrated excellent generalization and transferability across varying sample sets and imaging conditions.
• The developed system significantly accelerates stomatal density measurement.

Conclusions:

• The automated stomata detection model is a robust and transferable tool for oil palm research.
• This technology will expedite the breeding of oil palms with enhanced drought tolerance.
• Accelerated breeding selection is vital for sustainable agriculture in water-scarce environments.