Use of Computer Vision to Improve Construction Safety

Project Reference :


Institution :

National University of Singapore (NUS)

Principal Investigator :

Professor Goh Yang Miang

Technology Readiness :

9 (Actual system proven in operational environment)

Technology Categories :

Computer Vision

Background/Problem Statement

The computer vision market is expected to grow from USD 10.9 billion in 2019 to USD 17.4 billion by 2024-growing at a CAGR of 7.8% during the forecast period. Major factors driving the market growth include the increasing need for quality inspection and automation, growing demand for vision-guided robotic systems and rising demand for application-specific computer vision systems.

The construction industry is the top contributor of workplace fatalities. Currently, the industry relies heavily on manual supervision to prevent accidents. However, not only is manual supervision resource intensive, different site personnel have varying standards and practices in addressing safety risks, and quality of supervision will be affected by human physical and mental fatigue. Fall from open (unbarricaded) edges, and falling objects during crane lifting are common causes of severe injuries and fatalities on construction sites. Even though CCTV cameras are installed on site and cameras are mounted in tower cranes, the video footages are not actively reviewed to monitor workers and site conditions.


The missing object detection (MODA) approach is a vision-based approach developed based on the requirement to overcome the problem of detection errors arising from the shaking of cameras mounted in tower cranes and to facilitate the accumulation of training datasets. 

The key idea of the MODA method is to treat the ‘missing barricade’ as a type of object and exploit deep learning methods to directly detect the object in the image. MODA can achieve higher accuracy in the detection of missing barricades in images and has a higher detection speed than the masks comparison approach. The average precision and average recall for MODA were 57.9% and 73.6%, respectively.

(Computer vision approaches for detecting missing barricades)


  • By incorporating the technology with CCTV cameras, the construction site’s CCTV surveillance system can actively look for dangers on the site and alert the site. This reduces reliance on human supervision and such methods will not be limited by the working hours, vision, and attention of the safety supervisor. This increases productivity and safety at the construction site, and lower labor cost.  
  • Different safety supervisors have varying standards and practices in addressing safety risks. The use of the technology can encourage more consistent and more accurate standards and practices in automatically supervising the site. 
  • The unique algorithm used helps in detecting the 3-dimensional position of the area under a lifted load from a 2-dimensional image, which is not easy to do as the 2-dimensional image does not give information on the 3-dimensional space directly.
  • The algorithm is also unique in identifying where there are open edges (not barricaded) on the construction floor.

Potential Application(s)

This invention can potentially help to reduce the labor cost that is allocated to monitoring the safety of the workers and improve the productivity of the site. Currently, the sites employ the help of safety officers to ensure that the workers are complying with the safety standards set. Hence, with this invention, it is possible to lower the number of safety officers allocated to monitoring the safety of the workers and allocate other tasks to them instead. This will increase the productivity of the site and lower the labor cost while enhancing the safety of the workers at the site.

In the construction industry, computer vision has immense potential. Thanks to its object identification and recognition ability, it can assess video data from work sites in real-time, identify poor craftmanship, deviation from standardized work plans, or compare work done against BIM specifications. In terms of safety, it can monitor security camera footage and detect hardhats, high visibility vests, work goggles, shoes, and even special protection belts required for workers working at high altitudes.

We welcome interest from the industry for collaboration/ co-development / customisation of the technology into a new product or service. If you have any enquiries or are keen to collaborate, please contact us.