Case study: flame detection through vision
Return to showcase
With an increasing base of CCTV devices in the western world, a ready market is growing for more sophisticated camera applications. Sentec has used their understanding of image analysis to add value in a range of applications.
The opportunity
In our first project in the field, we worked with a leading provider of systems for detecting smoke in video. Conventional fire detection systems are ineffective in large open buildings or outdoors. For safety critical locations an alternative is to use image analysis on camera systems that spot threatening events, which reduces the need for labour intensive surveillance and the resulting operator errors. Fewer security staff are needed and their efforts can be focussed on responding to risks.
Our partners wanted to establish intellectual property in flame detection to work alongside their existing smoke detection algorithm, allowing them to offer a product that detected flame and smoke together, covering all the signs of fire
Our approach
There were several technical options to meet the product requirements. Our solution was to isolate the characteristic behaviours of flame in a 'skinny' algorithm that would allow the image processing to be done on the camera if necessary.
The system has been successfully trialled by our client with key customers in a number of sensitive environments.
Future developments
Sentec have since developed algorithms that can be run 'on camera' for detecting other interesting events:
- Intruder alerts
- Occupancy sensing
- Queue monitoring
These algorithms should have a number of benefits:
- Higher levels of operator productivity
- Higher levels of operator reliability
- Reduction in false alarms to authorities
We are discussing other applications and benefits of such 'skinny' algorithms with a number of clients, and are open to further ideas.
