Real-Time Detection of Strawberry Ripeness Using Augmented Reality and Deep Learning

Area of Science:

• Agricultural Technology
• Computer Vision
• Robotics

Background:

• Strawberry harvesting relies on manual labor and subjective ripeness assessments, leading to variable post-harvest quality.
• Automating ripeness detection is crucial for enhancing efficiency and consistency in strawberry production.

Purpose of the Study:

• To develop an automated system for accurate and efficient strawberry ripeness assessment.
• To integrate object detection and augmented reality for real-time visual ripeness feedback.

Main Methods:

• Utilized YOLOv7 object detection with transfer learning, fine-tuning, and multi-scale training for strawberry ripeness identification.
• Implemented augmented reality (Microsoft HoloLens 2) to overlay ripeness labels onto strawberries in real-world conditions.
• Evaluated model performance using metrics such as mAP and F1 score, and measured detection speed.

Main Results:

• The YOLOv7 model achieved high accuracy in detecting strawberry ripeness, with an mAP of 0.89 and an F1 score of 0.92.
• Real-time detection was achieved with an average detection time of 18 ms per frame at 1280x720 resolution.
• Demonstrated superior performance compared to other state-of-the-art methodologies.

Conclusions:

• The developed system offers a significant advancement over traditional methods for strawberry ripeness assessment.
• Augmented reality integration provides practical visual assistance for farmers during harvesting.
• This technology holds substantial potential for transforming agricultural practices and improving crop management.