background image
d e t e k t o r i n t e r n a t i o n a l · 2 3
Security News Every Day ­
www. securityworldhotel.com
Security News Every Day ­
www. securityworldhotel.com
video surveillance ­
market trends
Trend 6:
Embedded vision and
convergence with the wider industry
Niall Jenkins, Research Manager, IHS Technology
Embedded vision is not a new
concept to the video surveillance
industry. In fact, embedded vision
technology, in the form of video
content analysis (VCA), has been
used in perimeter-protection
applications for decades. More
recently, wrong-way detection,
people counting, and left-baggage
algorithms have all been deployed
in security and business intelli-
gence applications.
Embedded vision is the combi-
nation of embedded systems and
computer vision. Ultimately, it is
a technology that allows devices
to understand the world around
them through video; recognising
hand gestures and faces, as well
as interpreting behaviour and
actions. Embedded vision can be
broadly classified into augmented
reality, behaviour recognition, face
recognition, gesture recognition,
and object recognition applica-
tions. Almost 5 million device
shipments are estimated in the
established markets of embedded
vision in 2014. Video surveillance
represents only around 10 percent
of this number.
Historically, video content
analysis has also represented only
around 10 percent of embedded
vision device shipments, limited
to behaviour-, object- and face-
recognition algorithms. The sup-
plier base has likewise remained
relatively segregated from other
embedded vision markets in in-
dustrial automation, automotive
and consumer sectors. That is,
until now.
At the Security China 2014
show in Beijing, Gosuncn, a
Chinese systems integrator,
presented an "augmented reality"
camera. Augmented reality sup-
plements the user's view of the real
world with computer- generated
images. In the smartphone and
automotive market, augmented
reality can superimpose directions
onto a handset or the head-up
display in an automobile, making
navigation around a city much
easier. Augmented reality can also
make virtual characters appear to
interact with the real world on a
gaming handset.
The augmented reality network
PTZ dome camera is controlled by
touch screen, with objects in the
view tagged with overlays, show-
ing longitude/latitude, distance
and orientation. These labels are
shown on different layers on the
screen and can be switched on or
off, or selected to show a specific
category (e.g., restaurants). The
labels also appear and disappear as
the camera zooms into the image,
in the same way that street names
do in Google Maps. Furthermore,
users can input a street name or
click a label, and the analytics will
direct the cameras to locate and
zoom on that object; recording
can be triggered by touch screen.
This functionality is embedded
in the camera, and a Geographic
Information System (GIS) map
onscreen can show each camera's
orientation.
As we move into 2015, deve-
lopments in markets such as Ad-
vanced Driver Assistance Systems
(ADAS) should also filter through
to other embedded vision markets.
In automotive, there is a push to-
ward autonomous vehicles where
the embedded vision requirements
are much higher, as any false
decision could result in death.
This technology will also have to
analyse the foreground and the
background ­ both of which are
moving and may be poorly lit ­ to
recognise objects and behaviour, as
well as to make decisions.
The automotive market has
very strict qualifications and test
requirements that a product needs
to meet before it can be used on
a car. While this level of accuracy
will be expensive to develop, the
potential value of ADAS systems
and the self-driving car is likely
to see more money spent on re-
search and development than in
the other established markets of
security, industrial automation
and business intelligence. These
adjacent markets should benefit
from automotive developments
over the medium-to-long term.
Embedded vision
can be broadly clas-
sified into augment-
ed reality, behavi-
our recognition,
face recognition,
gesture recognition,
and object recogni-
tion applications.
Business & Product News
­ Every Day!
www.securityworldhotel.com