In spite of many jurisdictions
around the world having
introduced laws banning the
use of cell phones while driving,
the curse of distracted drivers
being glued to their hand-held
devices is one that persists.
The problem for authorities
is that detecting and proving
the offence with any regularity
has been a major sticking point.
That is until now.
Roadmetric is utilizing
advanced neural networks to
develop the first commerciallyavailable,
video-based
automated detector.
The software is capable
of detecting not just the
smartphone itself but, crucially,
also the typical body postures
of drivers using the device. Such
is the advanced nature of the
system, offences can be picked
up even when the phone itself
is obscured from the camera
by the individual themselves or
the vehicle they’re in.
Recording as soon as an
offence is detected, the smart
| Distraction Detection
camera captures up to 60
frames per second, with zoom
capabilities available as well –
making it a virtually watertight
case for a traffic court judge
to consider.
Having already used video
analytics to detect speeding,
parking and other common
traffic offences, Roadmetric is
confident the new software will
encourage greater compliance
with the laws of the road.
You can read more about this
system on page 60.
AI detection
Spotting drivers on mobile phones with AI has proved to be a technical
challenge, but one company has a commercially available system
Kinnear identifies four types of
distraction: visual, cognitive,
auditory and manual. Though visual
distraction (looking away from the
road) is the most dangerous, using a
phone can combine three or four
types of distraction, creating
culminative risks which increase
with the length of exposure. “Drivers
often get away with these things,
because nothing happens during the
distracting activity,” says Kinnear.
“But if distraction coincides with an
unexpected event, then it is more
likely to turn into a critical event,
usually with a crash as its outcome.
“There may ultimately be
a technological solution to this
technological problem, so that better
systems are developed to promote
good practice in the vehicle,” Kinnear
speculates. “It’s possible that Android
Auto and Apple Carplay are a step
towards disengaging the driver from
certain activities on the phone, like
social media, though these still
currently allow text-style
communication.”
The Ridy gadget, released last
year, uses a windshield-mounted
camera to analyze eye movement
and facial patterns, alerting the
driver if they begin falling asleep
or looking at their phone. Police
may soon be able to plug a
‘Textalyser’ device into a suspect’s
phone to scan recent activity
following an accident, just as
breathalysers are used to check
alcohol levels. But for policing
phone use in general traffic, basic
methods still prevail. NHTSA
reports that, in the absence of
credible technologies for
apprehending distracted drivers,
most US enforcement relies on
stationing officers at the roadside,
over high-volume intersections
Above: Roadmetric’s
commerciallyavailable,
videobased
automated
detector can identify
mobile phone use
with AI
July/August 2019 Traffic Technology International
www.TrafficTechnologyToday.com
025
/www.TrafficTechnologyToday.com