AI-Based Safety Recognition System for Steel Structure Construction

This project focuses on enhancing AI-driven safety recognition systems in construction, particularly for steel structure operations. It aims to improve existing recognition models, evaluate system adaptability across different construction environments, and address data drift challenges in AI recognition. The research involves upgrading AI-based image recognition systems to detect fall prevention facilities and personal protective equipment (PPE), ensuring an average accuracy of over 90%. Additionally, deep reinforcement learning models will be developed to improve adaptability, along with a true/false alarm feedback system to enhance recognition accuracy. To address data drift issues, field tests will be conducted at five different construction sites, analyzing recognition performance and proposing solutions to improve AI adaptability.

The project also explores category incremental learning by integrating new object detection categories, further expanding the AI model’s flexibility. By integrating advanced AI technology with edge computing, this research seeks to create a more precise, efficient, and scalable safety recognition system, ultimately reducing fall-related accidents and improving workplace safety in steel structure construction.