Exclusive Interview with Sören Wittman: AI in Fire Detection

Exclusive Interview with Sören Wittman: AI in Fire Detection

We had the opportunity to get an exclusive interview with Sören Wittman, Bosch‘s Production Manager regarding how Artificial Intelligence is to be used in Fire Detection in the future of the Fire Marketplace.

Find the full interview below:

Whether in the home, a production facility, or nature, a raging fire will always put people, property, and the environment at risk. There is a good reason why fire safety technology is heavily regulated. Does that heavy regulation leave enough scope for manufacturers and installers to explore innovative approaches and pursue new developments?

It is, of course, true that fire safety technology is very heavily regulated. That’s the way it should be. Ultimately, it’s about keeping people and property as safe as possible and doing that based on verifiable technical standards. Of course, we do sometimes wish that these technical standards could be drawn up a little more quickly, particularly when it comes to highly innovative technologies such as our video-based fire detection system Aviotec. As a manufacturer, we work closely with key institutions – such as VdS in Germany – so that we can extend technical standards in Europe and abroad bit by bit.

Smoke alarms are an established solution in the market. Using a video camera for fire safety purposes requires much more willingness to invest. What are the benefits you can expect in return?

When you’re trying to keep large areas safe, as is the case in industry and in warehouses, it doesn’t take long for individual, point-type smoke detectors to reach the limits of what they can do.

All the fire detection technologies that have been available till now depend on the smoke rising to their installation position. The higher the ceiling is, the longer it will be before the alarm will sound. If you also factor in the effect of air currents caused by ventilation, climate-control systems, and natural airflow, then it becomes really difficult to detect fires quickly.

Things are very different when using video-based fire detection. Smoke or flames that occur within the visual range of the fire detection camera become visible in the image almost instantaneously. The camera can detect the fire at its source in a matter of seconds. That saves valuable time in the response chain and this early detection can help minimize damage. This is an incalculable advantage when you’re talking about the fire safety of storage areas or expensive machinery that is essential for an organization’s very survival.

Additional benefits can be achieved by combining video-based fire detection with conventional video technology. Depending on what customers are looking for, that can open the door to a whole range of possibilities. For example, you can put alarm recordings in place. That involves video sequences being recorded in a ring memory and automatically deleted after a brief predetermined period. Data is only archived if an alarm occurs, which means the video material can be subsequently analyzed and potentially used to investigate the cause of the fire. Findings can also be utilized to help develop countermeasures so fires caused by similar circumstances can be suppressed in the future.

Does the software also help avoid false alarms at the same time?

Yes, thanks to the design and the combination of different algorithms, smoke and flames can be clearly distinguished from other objects in an image. We are doing our utmost to make detection as accurate as possible.

If an alarm is triggered, then video-based fire detection has another invaluable advantage. When a display device is being used, for example in a company’s security office, then the alarms can be quickly and reliably verified by an authorized individual looking at a live video. That person can very quickly establish whether it’s a false alarm or a genuine incident.

How will neural networks affect the future of smoke-detection solutions?

Deep learning is going to revolutionize fire safety, just as it will many other areas. We have been carrying out research in this field for years, including for video-based fire detection. It’s important to note that the detection rates, speeds, and even the robustness of fire detection can benefit from this new technology to an unbelievable extent.