AI integration in construction safety: Current state, challenges, and future opportunities in text, vision, and audio based applications
Ahmed Bin Kabir Rabbi,Idris Jeelani
DOI: https://doi.org/10.1016/j.autcon.2024.105443
IF: 10.3
2024-05-10
Automation in Construction
Abstract:High occupational injury and fatality rate in the construction industry is a serious global concern. Recognizing AI as a solution to enhance safety performance, this study reviews 153 papers to assess and categorize current AI applications in construction, focusing on text, visual, and audio data, while also identifying challenges and future research opportunities. Real-time monitoring, hazard detection, and information extraction are identified as key areas where AI is applied, with a notable reliance on deep neural networks, object recognition, and Natural Language Processing. The review highlights major challenges, including the need for high-quality data management, semantic feature representation, and occluded object detection. Additionally, it underscores the untapped potential of audio-based AI and the advancements possible with Large Language Models for text interpretation. The findings emphasize the need for integrated, multi-faceted AI systems and advocate for responsible AI deployment to mitigate safety risks on construction sites.
construction & building technology,engineering, civil