The artificial intelligence began to be used for forecasting of wildfires

In October, 2019 the CrowdAI company submitted the FireNet program which uses artificial intelligence for forecasting of wildfires.

The algorithm analyzes the images arriving from digital UAVs which turn over arrays of trees and take the picture with a speed of 20 frames per second.

Data processing happens in real time. Places with a high probability of ignition are determined by drone GPS coordinates.

CrowdAI submitted the FireNet program which uses artificial intelligence for forecasting of wildfires

According to developers, for training of artificial intelligence tens of thousands of the frames from videos with the fires were used. Thanks to machine learning the program can define borders of the fires with an accuracy of 92%.

CrowdAI warns that it is not necessary to apply FireNet as the only way to tracking the fires — the tool is developed only as addition and reduction of handmade amount.

But the artificial intelligence defines the fires much quicker, than people who at first watch video, comparing it to the card, and then report about it to operational services, noted in the company.

By October, 2019 FireNet is used by the Air Force of National guard of the USA.

Earlier in 2019 in one of the Belarusian forestry and landscape organizations the system of early detection of wildfires on the basis of intellectual video surveillance began to be tested.

The solution using neural network detectors of smoke includes three rotary cameras combined with thermal imagers and installed on fire towers at the height of 35 meters.

Cameras automatically rotate around a vertical axis by 360 degrees, fixing the smoke rising over the wood. The neural network which is specially trained at similar stages defines availability of smoke, distinguishing it from clouds, fog or dust. In good weather the range of detection can reach 50 km.^[1]

Robotics

Notes