Intelligent Hazard Detection (i-HaD)

About the project

Intelligent Hazard Detection (i-HaD) system is developed to exploit onsite images captured by CCTV surveillance cameras or site personnel to automatically identify and predict if physical health or safety of personnel at site is, or likely to be at risk. This project evaluates i-HaD by focusing on one of the popular construction hazards, that is the construction dust which leads to severe health issues especially on the long-term. Dust generated by many construction activities has been a persisting threat to physical health of construction workers for decades.

To automate the identification of dust generated at construction site, an object detection method is used with the help of a Convolutional Neural Network (CNN) and faster R-CNN in order to detect dust generating tools (e.g.: cut-off saws, grinders and drillers) as well as the workers’ PPE (e.g.: respirator) and material involved (e.g.: concrete) from an input of dataset of construction site images. The project presented a step towards the digital transformation and automation of the health and safety related tasks within the construction domain in order to achieve the ultimate goal of creating a safe environment for construction personnel.

The Challenge

Statistics and research have shown that construction industry is an extremely hazardous industry in terms of numbers of fatalities, permanent disablement cases and long-term health issues. While many inspection regulations, control measures and mitigation plans are implemented across the industry, they rely on limited human efforts. This includes the crucial step of identifying and detecting construction hazards at site which is essential for avoiding or mitigating hazards, and the effective health and safety management. While recent research highlights how large proportion of construction hazards remain unrecognized, the technological advancements in artificial Intelligence (AI), machine learning (ML) and image processing has brought innovative solutions with promising ways to supplement the existing construction health and safety practices, in particular the automated detection of onsite construction hazards.

The Solution

i-HaD uses an object detection method, with the help of a Convolutional Neural Network (CNN) and faster R-CNN, to detect dust generating tools (e.g.: cut-off saws, grinders and drillers) in addition to workers’ PPE (e.g.: respirator) and material involved (e.g.: concrete) from an input of dataset of construction site images.

Check out the image gallery below to see examples of what i-HaD can achieve.

Hazard detection example, person Drilling ground

Hazerd detection example, stihl saw with dust.

Hazard detection example, person using an angle grinder.

Hazard detection example, person using a stihl saw

Hazard detection example, person breaking up ground with drill

Hazard detection example, person starting a stihl saw.

The Impact

This project is a step towards the digital transformation and automation of the health and safety related tasks within the construction domain in order to achieve the ultimate goal of creating a safe environment for construction personnel.

Next steps, and future opportunities

The next step includes developing a reasoner as a part of i-HaD that is initiated to confront the onsite facts, including detected objects and the relations between them, with standard construction health and safety regulations. Consequently, i-HaD produces a list of classified images, from highly hazardous to non-hazardous ones. The produced list is ranked, as well as couched in terms of an existing formal classification of construction hazards.

External Partners: Arcus Consulting LLP

Research Dates: March – July 2022