Computer vision-based safe distance-estimation method to prevent forklift tip-overs

Area of Science:

• Engineering
• Computer Science
• Occupational Safety

Background:

• Forklift tip-overs pose significant risks in construction due to improper load positioning.
• Manual safety inspections are inefficient and require continuous human oversight.

Purpose of the Study:

• To develop a computer vision-based system for early detection of forklift tip-over hazards.
• To measure the critical distance between forklift forks and the load's center of gravity automatically.

Main Methods:

• Utilized computer vision with YOLOv8 for object detection.
• Developed a safe distance-estimation module (SafeDEM) integrating forklift length and safety regulations.
• Classified behaviors based on US OSHA standards.

Main Results:

• Achieved 93% classification accuracy in identifying safe/unsafe scenarios.
• The system demonstrated real-time processing at 24 FPS.
• Successfully tested with a 24-inch safety threshold.

Conclusions:

• The proposed method offers a scalable, non-intrusive solution for enhancing forklift stability.
• Enables early hazard detection before forklift movement, improving workplace safety.
• Supports future dynamic deployment in addition to static operations.