Temperature screening has become one of the main ways to identify someone infected with COVID-19 as high temperature is one of the main symptoms of infection, therefore monitoring temperatures is a key defensive tool in helping curb the spread of the virus.
GovTech Singapore has developed three systems for temperature monitoring in different situations.
Self-service temperature scanner
The first system – an automated self-service temperature scanner – was developed by the Sensors and Internet of Things (SIOT) team using off-the-shelf products, and the first to be publicly deployed in government buildings and community facilities.
The team started work back in February to re-engineer a non-contact infrared thermometer so that it would no longer require human touch to trigger the temperature-taking function. They aimed to create a device which took a temperature reading automatically when a person stands in front of it, at an appropriate distance.
The purpose of the project was to speed up temperature screening in crowded areas and reduce the number of human operators required.
“At a time of urgent need when people were all queuing up to take temperature just to enter their workplaces or buildings, it would have taken too long to put out a tender and invite vendors to manufacture these devices. Within GovTech, we have been building up our engineering capabilities for years, and this is the time where we put it to good use. We took a system from idea to design, prototype and production in just a week.” said Mr Quek Yang Boon, director of the SIOT team at GovTech when asked why GovTech was building the self-service temperature scanners internally.
VigilantGantry
GovTech’s Data Science and Artificial Intelligence Division (DSAID)developed VigilantGantry, a fully automated, contactless gantry system for temperature screening.
VigilantGantry works like the gates to an MRT station to ensure minimal physical contact. Rather than swiping your card to enter, a machine screens the temperature and takes a picture before letting anyone through.
GovTech has developed VigilantGantry, a video analytics model to identify cases missed out by current thermal systems such as fringe, cap, head-gear, and occlusion, using face and neck segmentation and colour segmentation algorithms. An alert will sound once the system detects an abnormal temperature, and entry will be denied, allowing for further checks on the visitor.
Existing thermal systems usually need at least two or three people per station and this is where VigilantGantry can create a significant reduction in the human resources deployed at public spaces, which is especially important during a pandemic.
While solutions like the self-service temperature scanner which uses a regular infrared thermometer might work in non-crowded areas, they are not feasible in places that see large crowds, said Mr Corey Chong, a manager at GovTech who is also part of the DSAID team.
“VigilantGantry was designed for sites experiencing a medium to high volume of human traffic,” said Mr Chong.
During the development process, the team faced many site-specific challenges, like making sure that the software could integrate with different thermal sensor models, space constraints and site regulations.
“The challenge was making our solution flexible, which meant doing a lot of iterative changes so that we meet the stakeholders’ needs,” said Mr Delon Leonard, an AI engineer at GovTech.
To achieve flexibility, the DSAID team aimed to make every component in VigilantGantry modular so it could be programmed with additional functionality.
Other than measuring body temperature and performing facial indexing, the system can also capture the location, date and time to support contact tracing purposes. It is also possible to store health and travel declaration data obtained via questionnaires.
VigilantGantry’s flexibility means that it can be scaled up easily, Mr Leonard added. For example, it can be integrated into existing thermal systems without the need to purchase new equipment, a bonus in these uncertain times when demands for thermal sensors are on the rise.
SPOTON
The Smart Nation Platform Solutions team’s SPOTON system does not need specialist hardware; it can use commercial off-the-shelf hardware such as a thermal sensor and a webcam mounted on a 3D-printed case. This case is then hooked to a tripod that can be adjusted to an optimal height for detecting human body temperature. Unlike other detection systems that require permanent fixtures at certain locations, it is mobile and can be easily set up.
“Our solution is focused on three things: being able to do crowd temperature screening, real-time screening and also screening at a safe distance,” said Mr Chong Jiayi, the engineer who led the Smart Nation Platform Solutions team in the project.
“It’s the software that’s doing the magic,” said Mr Chong Jiayi as he explained the inner workings of the temperature screening device.
Footage from both the thermal sensor and webcam is used to train the software to contextualise the information captured by both sensors, he explained. Once trained, the artificial intelligence -AI, can then automatically measure the temperature of the faces that come into its view. If anyone with a fever enters the frame, the device alerts the person-in-charge.
“This solution is unique because it uses a hybrid hardware and software, AI-driven solution. I don’t think there’s many of those in the market today,” added Mr Chong Jiayi.