| Technology Profile
Far Left: ALPR can be used
to reduce congestion
Left: Jenoptik’s system
relies on deep learning
system is trained once in the lab
and does not change once it has
undergone thorough testing and
is released for use on the roads.
Another example of
Jenoptik’s application of deep
learning technology is vehicle
classification. This is able to
determine the vehicle’s class,
such as car, motorcycle or truck,
directly from the photograph
taken by the ALPR camera,
and hence automatically charge
different amounts for each type
of vehicle. This system is also
based on a large set of example
vehicle images for which the
correct class is known.
Unlike traditional computer
vision algorithms, which rely
on researchers to provide the
sequence of computational
steps, the deep learning training
process optimizes its rules
automatically and is therefore
able to achieve superior
performance in difficult
conditions, such as night. Similar
deep learning networks can
also be used to determine the
vehicle’s color, make and model.
Based on its unique sensorfusion
solution, its ALPR and
deep learning expertise, the
Jenoptik RUC solutions achieve
the highest industry KPIs for
detection, measurement and
classification of vehicles.
taking a photographic still image
or video clips of the vehicle.
Jenoptik ANPR cameras can read
license plates from all around the
world. If the appropriate road
user charge has been paid for,
that particular vehicle’s details
are deleted. ALPR systems can
be fixed, vehicle-mounted or
portable. This increases the
effectiveness of the applications
the technology can support. The
flexibility of ALPR means both
proactive and reactive
monitoring strategies can be
implemented. ALPR – sometimes
combined with DSRC – can be
used in many areas where
satellite signals are either
disturbed or blocked.
processing, leveraging large
amounts of examples to ‘train’
a system to produce correct
answers without explicitly
programming the necessary
steps. Carefully curated training
data associates a diverse set of
possible inputs with their
corresponding required answers.
The deep learning technology
can then be applied to produce
very accurate responses for other
inputs of a similar nature.
For example, several hundred
thousand images of European
license plates from over 30
countries have been used to train
the Jenoptik deep learning ALPR
system. This patented integrated
method can reliably distinguish
what country the plate comes
from even when there are several
possible countries that allow the
same license plate syntax – it has
proven to be correct 99.8% of the
time. It is able to take into
account all aspects of the visual
appearance of the license plate,
such as character font, spacing
and any special markings such
as tax stickers. Because the
training images include realistic
levels of dirt, damage and
occlusion, it also learns to
operate well under all conditions.
It is important to note that the
September/October 2019 Traffic Technology International 071
www.TrafficTechnologyToday.com
For some applications, there
is no single sensor capable of
capturing all the information
needed in all conditions. The
best way is to combine sensors
and other components with
their specific advantages to
achieve the required results.
Jenoptik has modular multisensor
and infrastructure-based
object detection systems that
evaluate all information about
traffic conditions and road
users. Real-time measurement
data can be collected by the
ideal set of sensor technology –
stereo cameras, ALPR camera
with IR-illumination, radar and
laser – tailored to the respective
situation with its requirements.
Measurement accuracy
according to the position and
speed, the detection rate,
robustness to meet disturbances
like lighting conditions, rainfall
or snow and temperatures are
influenced by the capabilities of
different sensors with their
respective strengths.
Deep learning technology
The technology is underpinned
by collecting and interpreting
large amounts of information.
Deep learning is a cutting-edge
method of information
| Free reader
inquiry service
Jenoptik
To learn more about this advertiser, please
visit: www.magupdate.co.uk/ptti
/www.TrafficTechnologyToday.com
/ptti
