Construction Site Safety Management: A Computer Vision and Deep Learning Approach

Area of Science:

• Computer Vision
• Artificial Intelligence
• Occupational Safety

Background:

• Construction sites present significant safety risks, including falls and non-compliance with safety gear.
• Traditional safety monitoring methods can be labor-intensive and may miss critical incidents.

Purpose of the Study:

• To develop and validate image recognition models for improving construction worker safety.
• To assess the feasibility of using virtual environments for training and verifying AI safety models.

Main Methods:

• Developed three object recognition models: worker detection, fall risk assessment, and personal protective equipment (PPE) compliance.
• Utilized transfer learning with real construction site imagery and synthetic data from virtual environments.
• Employed virtual environments for efficient and safe simulation of accident scenarios for algorithm verification.

Main Results:

• Successfully created AI models capable of identifying workers, assessing fall hazards, and verifying helmet and vest usage.
• Demonstrated the effectiveness of virtual environments for generating training data and validating AI safety algorithms, particularly for rare events like falls.

Conclusions:

• Image recognition technology offers a promising approach to proactively enhance construction site safety.
• Virtual environment-based data generation and verification represent a significant advancement in developing robust AI safety solutions for hazardous industries.