Technology
SigniFire was inspired by a very simple
observation, that a visual
input provides sufficiently faster and more reliable information, then all
of our other senses combined. It
has been said "it's better to see once rather than hear a hundred
times". It has been our goal to develop artificial intelligence that will be able to
detect fires at the incipient stages of growth, using ordinary
television cameras. The guards, even if they are thousands miles away, are provided with an extra degree of confidence by live video images
that are sent from the location of the event.
Faster
Fire detection, for years, has relied on
smoke particles physically reaching the smoke detectors. Vision works
over larger distances at the "speed of light". This
fundamental difference alone may give vision-based systems tens of
minutes of lead time. Furthermore, due to a large number of false alarms, it became a standard
practice in many places to verify the event before dispatching the alarm to a
responder. With a vision system, an operator can verify the alarm
instantly as video images from the location pop-up on the screen.
Redundancy
SigniFire provides
a high level of redundancy by looking for many
different and unrelated aspects of fire, offering two distinct flame
detection methods and three smoke detection methods that include:
 |
Flame Detection
- Flame (pattern recognition)
- Off-site (flames are hidden behind the obstacles)
|
|
|
Smoke Detection
- Smoke plume (slow/fast)
- Smoke (diffusion effect)
|
In a typical situation the multiple detection methods
will all alarm during different stages of an event.
For example, during a flaming fire in the line of sight of the camera
the flame detection may trigger first, followed by the off-site
detection a few seconds later, then followed by the smoke detection.
During a smoldering fire the smoke detection may alarm first followed by
the flame detection and offsite detection if the smoldering fire were to
transition to a flaming fire.
Availability of vision input
The other advantage of a vision based technology is
the availability of the video data that becomes critical at
the time of alarm. Response personnel, using a real time video
feed are able to instantly assess the severity of a situation
and come up with most appropriate response. This information is most
useful when presented with other information (direction to the scene,
building contents, building floor plans, etc.) in a user-friendly format. Being able to see inside the burning building before
arriving at the fire scene may be
invaluable for a fire-fighting response team. The images will
provide a incident commander with the information he/she needs to
develop an appropriate response. Could a firefighter with a extinguisher
be sent in or should a defensive posture be taken. Being able to see behind the wall
may save the life of a firefighter entering a potentially dangerous
situation.
Born for DSP
All
SigniFire
detection algorithms are based on Digital Signal Processing (DSP) of
video images. It uses cascade of filters to convert visible images in
such a way that specific events of interest produce a signature
patterns. The DSP approach is ideal for implementation on the
embedded architectures that reside inside a video camera. |
