An artificial intelligence approach to real-time automatic smoke detection by unmanned aerial vehicles and forest observation systems

Abstract

Natural disaster events, like volcanic eruptions and earthquakes, usually cause serious havoc and result in property losses for millions of Euro and unfortunately sometimes in loss of human life. These events are very hard to predict and are out of human control. Much like them, wildfires can be devastating and can lead to significant destructions, but fortunately they can be detected, controlled and prevented from spreading. The evolution of the wildfire detection and prevention technologies is long. The traditional methods rely on human involvement and include the use of manned fire watchtowers and piloted fire observation aircrafts. These solutions are very effective, but they are also very costly and sometimes can be extremely dangerous. With the technological advancements in the different engineering areas, safer and cheaper solutions for early forest fire detection were slowly developed - fire detection satellites, fire and lightning detectors, wireless sensor networks, as well as static optical or thermal cameras. Nevertheless, all of these solutions have disadvantages, which calls for their constant improvement, as well as for the development of more advanced and reliable systems. In this paper we present our experience in the area of the systems for early forest fire and smoke detection. Following the introduction section, we analyse in details an artificial intelligence (AI) approach to real-time automatic smoke detection, which is suitable for use in camera-based static forest observation systems and unmanned aerial vehicles. The paper concludes with discussions on the capability of the presented approach and with directions for further work.