Robust real-time strawberry maturity detection using UAV-mounted deep learning for precision agriculture.

Area of Science:

• Agricultural Robotics
• Computer Vision in Agriculture
• Precision Agriculture Technologies

Background:

• Traditional greenhouse monitoring is labor-intensive and lacks scalability.
• Precision agriculture demands efficient, real-time monitoring solutions.
• Autonomous systems are crucial for advanced crop management.

Purpose of the Study:

• To develop an autonomous quadrotor UAV system for strawberry plant monitoring.
• To implement a robust strawberry detection model for greenhouse environments.
• To enhance precision agriculture through automated crop assessment.

Main Methods:

• Utilized a quadrotor UAV equipped with an onboard camera.
• Developed and tested a YOLOv9-GLEAN algorithm for mature strawberry detection.
• Implemented a hybrid trajectory tracking controller (PID+LQR) for UAV navigation.

Main Results:

• The YOLOv9-GLEAN model achieved high accuracy in detecting mature strawberries.
• The hybrid controller demonstrated superior tracking performance.
• The integrated system was effective in both simulated and real-world greenhouse settings.

Conclusions:

• Deep learning models, specifically YOLOv9-GLEAN, enable rapid, precise, automated ripe strawberry detection via quadrotor UAVs.
• This technology significantly surpasses conventional manual inspection methods.
• The system empowers informed decision-making to minimize crop loss and optimize yields.