Обнаружение лесных пожаров по изображениям, полученным с БПЛА

Abstract

Forest fires represent a serious natural risk causing economic losses, loss of life and significant environmental damage. In different countries of the world, including in Russia, research has been carried out to develop solutions for the detection and subsequent monitoring of forest fires. The most common ones are remote fire detection and monitoring systems, both groundbased (mobile or stationary), airborne (airplanes and helicopters), and satellite-based ones. These systems are highly efficient in data collection and fire detection, but for small areas. However, wildfires can cover large areas, making known approaches unsuitable for optimal spatial coverage. To overcome this limitation, unmanned aerial vehicles (UAVs) are proposed, that have proven to be useful due to their maneuverability, allowing the implementation of remote information acquisition in any direction, the flight route planning strategies, etc. Ultimately, they provide a low-cost alternative to established approaches for real-time detection and monitoring of wildfire areas. This article discusses a technique for early detection of forest fires based on UAV image processing, which is based on the previous works of the authors. The technique involves the use of computer vision algorithms for processing images received from an unmanned motor glider (UMG). The UMG route (for example, circular or elliptical) can be designed in such a way that it is possible to regularly receive information about a large area. The flight time of the UMG can be up to several hours. When transmitting the received images to the processing center, it becomes possible to extract information about the detection of a forest fire and subsequently control its spread.