Algorithm; Flame Detection Characteristics - Bosch AVIOTEC 8000i IR User Manual

AVIOTEC 8000i IR
3.2

Algorithm

The smoke and flame algorithm analyzes video frames by means of characteristic spatio-
temporal patterns and variables. The core of the detection algorithm is a Deep Learning
Network which was trained on a large-scale internal database, which is representative of the
use cases of a Video-based Fire Detection System. The fast detection algorithm is based on
a real-time image processing on the camera firmware.
There are factors that can influence this kind of visual fire detection. If possible, avoid
obstructions in the field of view, as they can have an influence on the detection speed in
case of fire. Flames behind obscurations cannot be detected and smoke needs to rise above
the obstruction to be detected in the field of view of the camera. In this case it is necessary
to analyze whether there is any need for further video-based fire detection cameras.
Division of the field of view into cells
The algorithm divides the image horizontally into 20 and vertically into 12 even grids to
analyze the video image. In each grid cell the algorithm decides if there is flame or smoke
visible in its area. There is a specific time span until the alarm triggers. This verification time
is a global counter: if at least one grid cell is in alarm, the verification time starts counting.
AVIOTEC 8000i IR can be used as primary fire detection system when the application is not
covered by any code of practice or other kind of standards or if the necessary approvals/
releases have been obtained for the intended applications in accordance with the applicable
laws and other regulations.
3.2.1

Flame detection characteristics

The Algorithm analyzes the video stream for potential fire events by searching for typical
flame behavior. To be detectable, a flame must be visible in front of the image background.
Transparent flames or flames with low luminance, like blue flames might not be detected.
Bosch Sicherheitssysteme GmbH
User manual
System overview | en 19
2023.10 | 02 | F.01U.414.168