Domain Feature Mapping with YOLOv7 for Automated Edge-Based Pallet Racking Inspections

Area of Science:

• Industrial Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Pallet racking systems are critical infrastructure in logistics and manufacturing.
• Manual rack inspections are costly, time-consuming, and prone to human error.
• Ensuring structural integrity of pallet racks is vital for operational safety and inventory protection.

Purpose of the Study:

• To develop an autonomous, computer vision-based framework for pallet racking inspection.
• To enhance the accuracy and efficiency of damage detection in industrial racking systems.
• To address data scarcity challenges in training machine learning models for this application.

Main Methods:

• Implementation of a computer vision framework utilizing the YOLOv7 architecture.
• Development of a domain variance modeling mechanism for synthetic data generation.
• Training and validation of the autonomous inspection system on pallet rack datasets.

Main Results:

• The proposed framework achieved a high mean average precision (mAP) of 91.1% in detecting pallet rack damage.
• Demonstrated the effectiveness of YOLOv7 for autonomous visual inspection tasks.
• Validated the domain variance modeling approach for improving model robustness with limited data.

Conclusions:

• Autonomous, computer vision-based inspection offers a significant improvement over manual methods for pallet racking.
• The developed framework enhances safety, reduces operational costs, and minimizes human error in warehouse inspections.
• This technology represents a step towards smarter, more efficient industrial operations.