SINGAPORE – More than 50 new Housing Board projects will have live video surveillance of their construction sites that will be analysed using artificial intelligence (AI) technology to flag potentially unsafe conditions or behaviour.
The AI system will analyse video footage and alert safety supervisors to high-risk situations that arise, enhancing worksite safety.
It will be integrated with closed-circuit television (CCTV) monitoring of construction sites, which the AI algorithm will sweep to detect potential hazards in real time.
HDB on May 27 said it requires the contractors of 22 new housing projects tendered from January 2023 to use such video analytics for worksite safety inspections. The firms are at various stages of adopting a suitable AI system.
Contractors of 30 housing projects tendered out before January 2023 will also have to roll out AI systems, as their works are still in the early stages and there is enough time for them to do so.
HDB said its adoption of such analytics comes ahead of new industry requirements by the Ministry of Manpower (MOM).
From June 1, video surveillance systems must be used at construction sites for projects valued at $5 million or more.
The maximum fine for workplace safety breaches will be raised from $20,000 to $50,000 from that date, said MOM on May 27.
This was among the measures announced by a multi-agency workplace safety task force in May 2023.
Senior Minister of State for Manpower Zaqy Mohamad, who visited Serangoon Polyclinic’s worksite in Upper Serangoon Road on May 27, said technology and automation will allow employers to detect safety breaches before they even happen.
“We want to create a safe environment and a safety excellence culture in Singapore such that it becomes the norm that employers look into the welfare of their workers, putting in place measures not because regulations need it but because they wish to do more,” he added.
HDB said the AI system that its contractors use must be able to analyse and detect high-risk situations such as open building edges without barricades and workers who come within 1m of them, workers who are near heavy vehicles, and those passing under a load being lifted who would be hit if it fell.
The AI system should be customised to detect other high-risk situations that are specific to each construction site, HDB added.
Once a safety hazard is detected, the AI system will immediately send a photograph of the scene with the location and time of event to the site safety supervisor’s mobile phone so steps can be taken to rectify the issue.
The AI system will analyse video footage and alert safety supervisors to high-risk situations that arise, enhancing worksite safety. PHOTO: AILYTICS
According to MOM’s 2022 and 2023 Workplace Safety and Health Report, common causes of worksite accidents include falls from height, incidents involving machinery and vehicles, and being struck by falling objects.
Noting that construction sites rely heavily on manual supervision, HDB said maintaining high standards of safety can be resource-intensive, needing multiple safety personnel to carry out checks.
In contrast, a video system that uses AI can carry out continuous and real-time monitoring of multiple sites with less manpower.
Safety personnel would be able to manage more areas within sites and monitor multiple sites at the same time.
The system would also reduce the risk of human error due to fatigue and inconsistent safety standards being applied by different personnel, HDB said.
HDB projects that require an AI system include Tanglin Halt Cascadia in Queenstown, which will have 975 units spread over four residential blocks, and Brickland Weave in Tengah, which will have 1,641 units across 12 residential blocks.
This AI system builds on HDB’s previous requirements for all Build-to-Order (BTO) projects tendering from 2017 to have CCTVs installed at construction sites to monitor construction progress.
HDB first tried AI systems at four BTO construction sites in Kallang-Whampoa, Clementi and Choa Chu Kang from 2019 to 2021.
It then expanded the pilot to eight more construction sites in Tengah in 2021 as part of an ongoing study that is set to conclude by the end of 2024.
As at mid-May 2024, there have been no fatal accidents at any of the sites where the AI system was implemented and no worksite accidents occurring in the high-risk situations identified by such systems on these sites, an HDB spokeswoman said in response to queries.
HDB has also been accumulating a repository of images involving elements commonly found at construction sites – such as workers, building, edges, machine-lifted loads, barricades, and precast modules.
This database will help AI systems learn to recognise objects at construction sites, and vendors offering AI services to HDB construction projects must share their collection of site images with HDB.
The images in HDB’s central repository can then be shared with industry partners to help other AI systems to build up the same capabilities.
AI technology has been deployed to enhance safety in other industries, such as transport.
Rail operators SMRT and SBS Transit have AI-powered video analytics systems that watch out for people falling onto the tracks at LRT stations.