Design and Development of a Public AI Referee Assistance System Based on Harmony OS Platform.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Sports Technology

Background:

• Commercial referee systems like Hawkeye have high costs, limiting public use.
• There is a need for accessible and accurate officiating solutions in sports, particularly for educational settings.

Purpose of the Study:

• To develop a small-scale, AI-powered referee assistance system for badminton.
• To provide fair and accurate judgments on controversial calls using computer vision.
• To enhance college students' participation in badminton training and competitions.

Main Methods:

• Utilized computer vision theory and deep learning object detection (YOLOV5).
• Employed computer graphics knowledge for judgment processing.
• Integrated the Harmony OS platform for a frontend mini-program.
• Implemented MediaPipe for accurate posture recognition.

Main Results:

• The object detection model achieved a prediction accuracy of up to 99%.
• The system accurately identifies in/out bounds and serving violations.
• The AI system demonstrates high accuracy in ruling and posture recognition.

Conclusions:

• The developed AI referee system is effective and accurate for badminton officiating.
• The system offers a cost-effective and accessible solution compared to commercial alternatives.
• It can help address the scarcity of referee resources and boost student sports engagement.