Working-at-high operation safety protection recognition based on target detection and spatial relationship

Area of Science:

• Construction Safety Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Construction workers face significant safety hazards, particularly falls from height (FFH), due to challenging tunnel environments.
• Lack of protective guardrails is a major contributor to FFH accidents in tunnel construction.

Purpose of the Study:

• To develop an automated system for detecting the safety protection status of workers operating at height in tunnels.
• To enhance on-site safety management and mitigate FFH risks.

Main Methods:

• Image preprocessing using Cycle-consistent Generative Adversarial Networks (CycleGAN) to improve image quality.
• Target detection of workers, trolleys, and guardrails using a YOLOv8 model with a coordinate attention (CA) module.
• Spatial relationship analysis for identifying safety protection status.

Main Results:

• The proposed computer vision model achieved high accuracy (95.89%) and recall (97.22%) in identifying safety protection status.
• The system effectively detects workers, trolleys, and guardrails in complex tunnel environments.

Conclusions:

• The developed method offers an intelligent solution for monitoring working-at-height safety in tunnels.
• This technology assists management in preventing FFH risks and enhancing overall construction site safety.