Engineers at the Moscow Aviation Institute (MAI) have developed an intelligent system capable of detecting fires at a distance of up to 5 kilometers in 5–7 seconds, accurately distinguishing smoke from steam and determining the geographic coordinates of the source. The solution, compatible with commercially available drones equipped with conventional cameras, significantly reduces the response time of emergency services to forest fires.
According to the project leader, Nikita Laletin, a student at MAI's Institute No. 3 "Control Systems, Informatics, and Electrical Power Engineering," the technology is based on a neural network model of computer vision. The model was trained on video footage provided by the Russian Ministry of Emergency Situations. As a result, the system achieved a smoke recognition accuracy of 95.1%.
Currently, the prototype has undergone testing in laboratory conditions. In 2027, the development team will conduct field tests and demonstration trials at the training grounds of potential customers, including specialized government agencies and environmental services.
