Adaptive Reflection Detection and Control Strategy of Pointer Meters Based on YOLOv5s

Area of Science:

• Robotics and Automation
• Computer Vision
• Artificial Intelligence

Background:

• Reflective phenomena in pointer meters hinder accurate readings by inspection robots in complex industrial settings.
• Existing methods struggle with real-time detection and mitigation of these reflective areas, impacting operational efficiency.

Purpose of the Study:

• To develop an adaptive deep learning-based method for detecting reflective areas on pointer meters.
• To implement a robot pose control strategy for eliminating detected reflections.
• To enhance the reliability and accuracy of pointer meter readings in challenging environments.

Main Methods:

• Utilized YOLOv5s (You Only Look Once v5-small) for real-time pointer meter detection and perspective transformation preprocessing.
• Improved the k-means clustering algorithm using YUV color space information for adaptive reflection detection.
• Developed a robot pose control strategy to adjust the robot's position and eliminate reflective areas.

Main Results:

• Achieved high detection accuracy of 0.809 for reflective areas.
• Demonstrated significantly reduced detection time of 0.6392 seconds.
• Experimental validation on a custom inspection robot platform confirmed the method's effectiveness.

Conclusions:

• The proposed method effectively detects and mitigates reflective areas on pointer meters, improving inspection robot reliability.
• Offers a robust solution for real-time reflection avoidance, crucial for complex industrial environments.
• Provides a valuable technical reference for developing advanced inspection robot systems.