07 February 2020

Tech Perspective #2: Put Out (Any) Fires using Algorithms


Wildfires, Bushfires, Forrest Fires – these are some of the most horrifying cataclysms faced by humanity. How can Machine Learning algorithms help?

  • AI tracksand predicts path of devastatingwildfires
  • Usingimage recognition through Machine Learning
  • Furtherpotential: urban planning,real estate,risk assessment

Wildfires, Bushfires, Forrest Fires – where these events occur, devastation follows. But there may just yet be a solution, with wide-ranging applications, using the power of algorithms.

Fires thrive in a variety of conditions, and their spread can be unpredictable, owing to an unlimited list of variables. This makes preemptive firefighting and prevention a near-impossible task, testing to the capacity and morale of firefighters across the world.

Predicting and Preventing Future Disaster

By what if we could predict the path of a fire before it occurs? It’s a simple case of machine learning, which has helped firefighters identify problems before their occurrence.

Through image analytics of satellite-based images of forest fires, systems can recognise the events to a 98% accuracy. Then, by mapping past movements of fires with current weather and wind activity, using historical composites of satellite images, machine learning helps predict the spread.

To take this a step further, by integrating such technology with the likes of communication tools, the system can provide instant and precise notifications to evacuee civilians and inbound emergency personnel, informing each group of where to go, and when. 

With Artificial Intelligence, predicting future natural disasters and their effects comes down to understanding the past. This way, fighters can contain these destructive phenomena before they cause any real damage.

So, the question is, why isn’t this technology used elsewhere?

Worldwide Application

There are multiple natural phenomena occurring, from coastal erosion, to flooding, to other kinds of terraforming. All of this could be monitored, measured, and accurately predicted!

The urban planner, whose job it is to map and develop new areas of cities, would assess the likelihood of terraforming in certain geographical areas. Coastal erosion is one key problem area in major cities.

Real estate valuations and insurance policies could become more intelligently estimated. Using predictive image-based analytics, insurers could pass on more accurate – and possibly more lenient – rates onto their clients. The primary beneficiaries could be those whose properties lie in traditionally risk-prone areas.

Agribusiness, which would likely merit from any increased predictability of crop and land output would likely experience a boost.

In fact, we can use image analytics to assess any vulnerable land or property. Even large infrastructure projects could cushion themselves from contingencies, with insights into flooding or the onset of terraforming.

Machine Learning and the Impact for Your Business

By training a system in image recognition, safety for millions of lives and properties can be ensured.

But the preventative measures take this a step further and have further applications for private enterprise too.

With the power of Machine Learning, putting out literal (or metaphorical) fires becomes that much simpler.

Read about business application of cutting-edge solutions!

Marcin Bartoszuk
Chief Operating Officer

With Microsoft technologies related since 2005. He graduated from the Computer Science Faculty of the Bialystok University of Technology where he was the leader of the .NET Group and the Microsoft Student Partner. Four times finalist of the national stage of the Imagine Cup competition, and later the mentor and the jury member of the contest. Co-founder of the Bialystok .NET Group. He lectured .NET development at the Bialystok University of Technology. Microsoft MVP in the Client Application Development category in 2008-2010, when he actively participated in the IT community. Constant new technology enthusiast and IT consultant.