Raymetrics - Early fire detection presentation

Raymetrics - Early fire detection presentation

2. Basic Info
The equipment is placed on a
special platform at a suitable
height (about 5-10 meters).
It rotates 360 degrees
horizontally and 90 degrees
vertically.
The movements of the system
depend on the mapping of each
detection location (ground
relief) and are different at each
installation point of the system.

3. Smoke Detection
•After we take a frame from the EO or the LWIR
camera, we use it as an input to an Object
Detection System (CNN algorithms).
•That will give us as an output, the answer if there
was smoke detected and if so, where it was
detected so our lidar can aim this spot and verify
in a very small amount of time if the detection is
correct, in order not to give a false alarm.
•Additionally, if there is smoke detected, our
object detection system provides the confidence
of the detection (very early smoke in 5 km
distance is detected with less confidence than
easy-to-see smoke in a near distance)

4. Input Frame from EO camera

5. Output Frame / Smoke-like object detected

6. Lidar
Verification
This RAY FD-LB101-ESS-D200 system
differs from all other conventional
camera systems because it
additionally has a specialized LIDAR,
so not only we can detect early smoke
from our neural network system, but
most significantly our system can
detect and confirm the existence of
forest fire particles using the lidar.
So, when our object detection system
detects smoke, we use our LIDAR to
verify if it is indeed smoke or we
have a possible false warning.

7. Lidar Verification

8. DIFFERENCE BETWEEN A CONVENTIONAL FOREST FIRE
DETECTION SYSTEM WITH ONLY CAMERAS AND THIS
SYSTEM THAT ADDITIONALLY INCLUDES LIDAR
Conventional systems:
•Many false alarms (one example is wrongly detect a cloud as smoke)
•Need lot of time to give an alarm ( because they have towait for early smoke to grow so
that the neural networks can have a high enough confidence >80% that they have
detected smoke, so they can call an alarm)
So, many false alarm and requires lot of time and bigger smoke (to have high confidence) to
call an alarm
Our system have:
•No false alarmsdue to the lidar verification (in the cloud example we find it out with
depolarization)
•Needs a little time to verify if there is smoke,because the moment our system detect
smoke even with low confidence (as real early smoke in 5 km distance covers too little
pixels to have high confidence >80%) we confirm the existence (or not existence) of
forest fire particles with our lidar.
So, zero false alarms and very quick lidar verification, even if very early smoke with low
confidence is detected

9. So, if there is smoke, even in a
very early stage of the fire,
our system will detect it with
our object detection system
and it will verify it with our
lidar, with no false alarms, in
a very little amount of time.

10. Detection
Node
The detection equipment
operates 24/7, completely
autonomously and routes all
information to the control
center via internet,
microlink, satellite or
4G/LTE.

11. Mobile Node
In addition to the fixed nodes there is also the solution of the mobile
node with the system installed on a specially configured trailer. The
mobile system can be placed temporarily at selected points to condense
the information, or it can also replace asystem that is currently not
working for maintenance reasons.

12. Control Center
•The control center can monitor an unlimited
number of nodes. Based on the information
received from each node, it is analyzed and
captured on the map of the area in real time.
•The map with various configurable colors and
symbols shows the probability of 0-100% that
there is fire in various points, as well as the
points checked by LIDAR.
•The information system of the control center
has the ability tomanage the potential alarms
as well as share the information on devices
(tablets, mobiles etc.) of others authenticated
users with relevant authorization.

13. Control Center Screen

14. Thank you for your
attention