| |
|
|
|
|
|
Technology
Humans are
very skilled at recognising situations visually. However, it has
always been a challenge to get computer-based systems to
perform such recognition with accuracy. Smart Sensor technology
uses an appearance-based approach to visual recognition and
achieves consistently good results. Real-time or stored images
are processed using task, scene, function and object contextual
knowledge to improve the recognition processes.
 |
 |
|
Modules |
|
|
There are several discrete stages in the
process. Sensors view a scene and key features
are extracted and grouped. Objects are then
identified, recognised and tracked both
spatially and temporally. Contexts and
situations can be learnt and events triggered if
specified parameters met.
The software is available in modules that
address specific situations; counting,
identifying gender, measuring
density, observing queues, measuring attention
time and tracking objects. The underlying core software can be adapted to
other situations if required.
Modules can be combined to further enhance
reporting possibilities. |
|
 |
 |
|
The visual advantage |
|
|
The use of visual image processing has a number
of advantages over other technologies such as
the use of lasers, infra-red sensors, thermal
imaging or mechanical detectors:
-
Video based systems are intuitive to set up
and administer.
-
They have the advantage of being suitable for
any ceiling height – ideal for large public
spaces and for outdoor situations.
-
By using commercially available video capture
equipment and processors, value is added in
software. This allows an evolutionary
development so that a system installed today
can continue to be upgraded with new versions
of software.
-
Video based data collection is very adaptable;
there is no requirement to modify a floor plan
or customer flow in order to accommodate
sensor limitations.
|
|
|
|
Indoors and outdoors
Smart Sensor technology is also suitable for outdoor deployment as in
traffic and crowd monitoring.
Linked sensors
Many sensors can be
linked together to form an observation matrix. Continuous tracking
is therefore possible within that matrix. |
|
|