To improve worker safety and reduce the need for manual labour, the Housing Board (HDB) will try out a surveillance system that automatically detects high-risk situations at construction sites and sends alerts to the person in charge.
HDB will try out the system at several Build-To-Order (BTO) worksites, including the first trial at Clementi Peaks which will start later this month. The trials are expected to be completed by the last quarter of next year.
The project, in collaboration with the National University of Singapore's School of Design and Environment with support from AI Singapore, is among the technological innovations HDB is working on to improve productivity and safety.
It shared details with the media in a briefing last Thursday.
There are two scenarios being tested by the system, which uses video analytics and artificial intelligence (AI).
One, when workers are within a metre of a non-barricaded building edge with a fall height of more than 2m; and another, when they are directly under the fall path of heavy loads lifted by tower cranes.
When such situations arise and are caught by closed-circuit television cameras, alerts will be sent to the safety supervisor's mobile phone via the Telegram app, allowing him to take action if there are lapses.
Such a surveillance system, to complement manual inspections, would enable round-the-clock supervision, and reduce human errors that occur due to fatigue, said HDB.
If the trial is successful, the tool could be used to enhance worksite safety in other areas, such as when workers are in the way of vehicles.
The workplace fatal injury rate for the construction sector rose to 3.1 per 100,000 employed persons last year, with 14 fatalities - two more than in 2017.
There were more than 20 deaths per year from 2014 to 2016.
The sector is among the highest contributor to major injuries in the workplace, with 124 cases last year.
Among the most common incidents leading to such injuries are falls from height and being struck by falling objects.
HDB is also looking into using drones with its partner H3Zoom.ai to inspect building facades, to improve worker safety and reduce dependence on manual labour.
Typically, workers in suspended gondolas inspect buildings. It may take several days to inspect one block manually.
In a pilot trial conducted in Yuhua and Sembawang from July to October last year, the drone inspections took at most a day to complete.
The drone inspection system taps a cloud software platform to visually scan building facades captured during inspection.
Using AI, the system can process thousands of photos to identify building defects and categorise them. The platform then delivers a report to highlight the severity of the defects and recommend possible remedies.
HDB said it will continue to refine the system to overcome limitations, such as when inspection involves entering building recesses and air-wells which could interfere with Global Positioning System signals.