Utilizing convolutional neural network (CNN) for orchard irrigation decision-making.

Area of Science:

• Agricultural Engineering
• Computer Science
• Plant Science

Background:

• Efficient agricultural management, especially in challenging terrains, is labor-intensive and relies on farmer experience.
• Smart technologies offer solutions to reduce labor burdens in agriculture.
• Precision irrigation is crucial for optimizing crop yields and resource use.

Purpose of the Study:

• To develop a decision-support tool for smart irrigation in orchard systems using a convolutional neural network (CNN).
• To assess the model's accuracy in identifying water stress in persimmon trees based on leaf images.
• To evaluate the potential of the CNN model for remote irrigation systems in mountainous regions.

Main Methods:

• Collected soil moisture data and leaf RGB images over two growing seasons for persimmon cultivation.
• Trained, tested, and validated a CNN model to determine optimal irrigation timing.
• Utilized image analysis to detect water stress levels in trees.

Main Results:

• The CNN model achieved over 80% accuracy in identifying water stress levels from persimmon leaf images.
• Demonstrated the potential of leaf image analysis for automated irrigation decisions.
• Highlighted the model's capability as a component of a remote irrigation system.

Conclusions:

• The developed CNN model shows promise for smart irrigation in persimmon orchards, particularly in challenging terrains.
• Further research is needed to improve model robustness and adaptability to diverse field conditions.
• AI-driven irrigation management can enhance agricultural efficiency and sustainability.