Improving construction safety through AI-enabled visual attention monitoring

Project details

Full project title: Improving construction safety through AI-enabled visual attention monitoring: An immersive approach project

Duration: 3 February 2025-2 February 2027

Funded by: Department for Science, Innovation and Technology

Project team: 

• University College London
• University of Tokyo

Partners:

• Balfour Beatty (UK)
• Yokogawa Bridge (Japan)

Project summary

Globally, the architecture, engineering and construction (AEC) industry accounts for over 100,000 fatalities annually. In industrialised nations, such as United Kingdom and Japan, construction workers are three to four times more likely than workers in other sectors to die from accidents at work. Apart from the psychological impact, the prevalence of accidents and fatalities in construction, pose a considerable amount of economic burden on society (estimated to be 4% of global GDP). While effective hazard recognition is one of the important and fundamental approaches to proactive accident prevention on construction sites, over 50% of construction hazards remain unrecognised.

One of the key underlying reasons for poor hazard perception is behavioural issues, often related to cognitive responses to danger and hazardous situations. Previous studies have attempted to address the inadequate onsite hazard recognition ability by developing numerous assistive tools. However, there has been less exploration of human augmentation technologies for improving visual perception and cognition in industrial scenarios. Human augmentation is an Industry 5.0 concept where individual capabilities and cognition can be amplified using technology. The broader vision of Industry 5.0 is to make industrial transformation more people-centric through technologies that simultaneously improve efficiency and interactions between the people, the physical and cyber environments they work in. Thus, it has become necessary to investigate hazard recognition to provide insights into how workers perceive construction sites, why hazards go unseen and how novel augmentation techniques can be leveraged.

Few studies have applied emerging Artificial Intelligence (AI)-enabled technologies to investigate the role of visual perception and cognition on hazard recognition performance, as well as the mediating effect of individual characteristics, cognitive psychological profiles, cultural and site-specific factors. Artificial Intelligence (AI) enabled eye-tracking is one key enabler of human augmentation and can help in understanding human visual attention and cognitive responses to risk in industrial settings. There is, however, limited understanding of how to characterise visual attention and its measurement for the purposes of hazard recognition.

This project, thus, proposes to explore visual attention and its measurement using immersive simulated environments. It aims to develop a novel Virtual Reality (VR) based application that integrates AI for real-time assessment of visual-search and gaze behaviour of construction workers during execution of tasks. The immersive simulation platform will enable characterisation of visual attention and its measurement in construction scenarios. The study will thus provide pathways for the development of Industry 5.0 human augmentation techniques leveraging real-time visual attention assessment for safety management. This approach will provide pathway for development of augmentation technologies for industrial safety in three areas: real-time alert systems; simulation for safety intervention planning; and personalisation of safety training leveraging visual attention and cognition feedback.