Facial recognition technology watches out for risks and monitors driver behavior
A staggering average of 66 people die in auto accidents every day in Thailand. Many smashes involve fatigued and reckless drivers who speed along highways that are choked with motorbikes, cars, trucks, and buses.
One of the most dangerous routes in the Southeast Asia kingdom is a congested 180-kilometer (111-mile) stretch between its capital, Bangkok, and the busy industrial hub of Rayong.
PTT Global Chemical Public Company Limited (GC) has offices and factories in both centers. Every year its employees clock up around 8,000 trips between the two locations in the company’s in-house fleet of cars and vans.
Knowing that corporate road travel was becoming a matter of life and death, Chatchawalit Dhammasaroj, GC ‘s Vice President of General Administration, searched far and wide for ways to reduce the risks.
He found it on the other side of the Pacific at Microsoft’s headquarters in Redmond in the U.S. state of Washington, where he saw facial recognition artificial intelligence (AI) in action for the first time.
Making the case for safety
With a good idea in place, Dhammasaroj faced his next business problem — budget. “My trip was past the budget planning time already, so I had an idea without a budget.”
But knowing how important this was, he raised the concept with management at a safety meeting, and they were very interested.
The next step was to work with FRONTIS, a leading consultancy in strategy and digital transformation, to estimate the costs, design and implement the project. A special budget was allocated and the program, AI for Road Safety, was born.
AI for Road Safety
The system, which is still being refined, combines facial recognition AI with video and data analytics to monitor a driver’s behavior in real time. It aims to detect when he or she is feeling drowsy or distracted.
It works like this: Company vehicles are fitted with cameras focused on the driver, as well as a GPS (Global Positioning System) to detect speeding. Facial recognition data is collected and transferred to the cloud where it is interpreted using machine learning.
Drivers will get an immediate alarm if they show signs of risk. And, the fleet manager can dispatch a relief driver, if necessary.
Management also gets a bigger longer-term picture via a dashboard supported by data analytics from Microsoft Power BI. This gives a fleet manager a range of options to help improve the well-being of their drivers and the people they ferry.
For example, stretches of road that are inherently more dangerous than others can be identified, so that drivers can be alerted, and routes altered. Fleet managers can examine details of past accidents to help prevent future ones. They would also be able to get a picture of individual drivers’ strengths and weaknesses, giving them a chance to improve performance and provide training, if necessary.
Dhammasaroj said, “we worked with our drivers to develop the solution and implement them in 4 vans. Our first consideration was that the data should be collected in accordance with high privacy standards and used purely to prevent accidents and retrain drivers.”
Permsak Pengprapa, shuttle driver with GC shared, “This is the first time I am encountering Artificial Intelligence in my every day work and I think it’s a positive change. I have been a lot more mindful while driving and continuously learning how to improve my driving skills.”
Several challenges arose as the system was being implemented. The first prototype used a video camera that was too bulky and obtrusive when installed in a van. This made drivers uncomfortable and unable to focus on the road ahead. That was easily overcome with sourcing a different, smaller camera.
Other problems were harder to deal with. The system uses Microsoft’s Cognitive Services Face API to recognize, identify, and read emotions from faces. According to Dhammasaroj, this is made difficult by different human behaviors.
“When we’re talking about detecting drowsiness, the number of behaviors can be high, so we are working to find the right number,” he explains.
As the system is used more, machine learning from the data collected will continually improve the detection of drowsiness and should even be able to recognize unique behavioral signals.
A holistic approach
Beyond the prime goal of employee safety, the system promises to improve the company’s performance in additional cost-saving ways.
Routes and schedules can be studied for inefficiencies and adjusted to better match people’s needs — and use fewer resources. Eventually, data analytics from the GPS can be used to assist with preventative maintenance and to look for ways to reduce fuel consumption by assessing speeds.
A first for Thailand
The AI for Road Safety program, using the Cognitive Services Face API, is a first-of-its-kind initiative for Thailand. It has become a point of pride for those involved and a matter of interest by authorities as it deals with the unfortunate danger of Thai roads.
GC sees the program as a major component of its intention to contribute to society, especially in Rayong, where the company has a number of factories. It plans to deploy the technology at 10 more subsidiary companies over the next three years, impacting more than 4,100 employees.
“Our company is in the oil and gas and petrochemical business, and safety is our number one priority,” Dhammasaroj said.
“Our ultimate goal is to have a zero-accident rate across our factories, offices, and plants. The impact of this goes beyond our company and extends to the families of our employees who at the end of the day just want their loved ones to return home safely. We believe this is achievable with the support of our people and technology.”
The company is also open to selling the safety solution to other organizations to reduce the overall road accident rate of Thailand.
Resource : Microsoft