|
3D taken for granted
Interestingly we have seen many recent
claims recently about products on the
market with a 3D or even 4D capability.
This sophisticated sounding term
essentially means that the product
can handle perspective (the fourth
dimension is time). Often this involves
using 2 cameras.
Almost 8 years ago iOmniscient
pioneered the technique of
understanding perspective within a
2D video image. This ability to handle
perspective with a SINGLE CAMERA
is available in all iQ products. We don't
make a big noise about it as it is one
of the simpler aspects of our system
which is capable of far more complex
and sophisticated analysis. |
|
Winner - Engineers
Australia Award – 2010
Adding to its long list of International
Awards, iOmniscient has won the
nationwide Engineers Australia Prize.
|
The Prize was awarded for
its implementation of the
Safety System on the China
Fast Train project. |
|
|
iOmniscient appoints
ClearView in the UK |
|
ClearView Communications is a
leading integrator of high-end security
equipment and systems. Based in
Chelmsford, Essex in the UK, they
design, install and support systems
around the world. Clearview has
many large customers. In the UK these
include Police, Government, Oil & Gas,
Education and many large commercial
organizations.
After extensive evaluation Clearview
has decided to standardise on
iOmniscient products and will be
exhibiting them at both HOSDB and
IFSEC. |
|
|
See us at
Intersec Trade Fair and Conference
16th- 18th January, 2011
Dubai, UAE |
|
|
India's Durga Puja protected by iOmniscient |
|
Every year millions of devotees congregate
in different parts of India in what is one of
their most colourful festivals. The primary
requirement in such an environment is managing
crowds.
Security officials need to understand
if
criminal elements have infiltrated the crowd,
if suspicious objects have been |
abandoned in crowded places and how many people there are
at a location at any time.
In 2010 iOmniscient's comprehensive video
analytics system which combines detection and
identification capabilities was used to provide
security at the Durga Puja. iOmniscient's system
was used as no other supplier can cope with Crowded Scenes. |
|
Understanding Facial Recognition |
As the first Facial Recognition systems for Crowded Places begin to get deployed in the
market by iOmniscient we have found a growing tendency from competitors to try and
obfuscate the differences that exist between traditional Face Recognition and iOmniscient's
Face Recognition in a Crowd. Here we will explain at a very high level, some of the key
differences. |
Traditional Facial
Recognition Systems |
Facial Recognition
in a Crowd |
• Used for Access Control. Usually involves
One to One or One to Many Recognitions. |
• Used for Surveillance. The person to be
identified may not be aware that his face is
being recognised. Many people may be walking
in the scene and all of them can be recognized
as they approach. |
• The person to be identified must be
co-operative and his picture must be taken
close-up with hi-res (eg megapixel) camera.
The resolution must provide at least 90
pixels between the eyes to achieve reasonable
accuracy and up to 300 pixels for very high
accuracy. The accuracy degrades very quickly
with lower resolutions. |
• The image may be taken from a distance
using standard security cameras that provide
an image with an average of 22 pixels between
the eyes. The accuracy will not deteriorate
calamitously as the resolution falls even to 12
pixels between the eyes. |
• The images in the database used for
comparison must be in high resolution and
taken in a controlled environment. |
• The images in the database may also have a
resolution that only rovides an average of 22
pixels between the eyes. |
• While accuracies in the high nineties are
possible alternative technologies such as
finger print or iris recognition can usually
provide higher accuracies where the person
is required to be co-operative. If Face
Recognition in this environment can provide
99% accuracy and the alternatives provide
99.9% which would you choose? |
• There are no alternative technologies for
recognising people in a crowd. To be useful
the system has to out-perform a human
attempting to do the recognition. |
|
|
|
|