Equipment Identification and Localization Method Based on Improved YOLOv5s Model for Production Line

Area of Science:

• Artificial Intelligence
• Computer Vision
• Industrial Automation

Background:

• Intelligent video surveillance is crucial for Industry 5.0, but current methods struggle with low recognition accuracy and localization precision for production line devices.
• Challenges persist in real-time monitoring and precise identification of equipment like robotic arms and AGV carts.

Purpose of the Study:

• To propose an enhanced YOLOv5s model for improved recognition and localization of production line devices.
• To address the limitations of existing intelligent monitoring systems in terms of accuracy and precision.

Main Methods:

• An improved YOLOv5s model incorporating a CA attention module, GSConv lightweight convolution, Slim-Neck, and a Decoupled Head structure was developed.
• The method focuses on real-time detection and precise localization of industrial equipment.

Main Results:

• The improved YOLOv5s model achieved 93.6% precision, 85.6% recall, and 91.8% mAP@0.5.
• Testing on the Pascal VOC2007 dataset demonstrated significant improvements in recognition accuracy.

Conclusions:

• The proposed method substantially enhances the intelligence level of production lines through accurate device recognition and localization.
• This research offers a valuable reference for manufacturing industries pursuing intelligent and digital transformation.