Firevision Lucile Rossi

Forest fires are a major hazard in many countries of the world. The University of Corsica has been working on this topic for fifteen years and there are currently 17 pesons working on this project. We experiment to better understand the phenomena that occur during propagations. To obtain information on the fires when they are propagating, we use vision by setting the cameras all around the field. They are in the visible range and the infra-red range and from every stereovision system we can reconstruct the fire in 3D and so to be able to extract the geometric characteristics like speed, location area and size. The overall form of fire from different cameras, is obtained by assembling the various 3D shapes catched by each of them. And to compile all these data, we have to use a model which has reference points for each of these cameras, and we can do it thanks to a cube. As far as I know there is no system identical to our own in the world. Today we are also working on the ground, because the aim of our work is the search and the modeling of forest fires. Here we are on a by 10m platform This sometimes causes problems when transferring equipment and so we have no other choice but to use a helicopter for that. The results are given in two formats, The first is an interface which for each instant shows the fires captured and the results we have obtained by stereo vision estimations. The second is an another interface too It copiles all the results we obtained, for the total length of the propagation. In the future we hope to develop a system based on stereo vision for the research and the modeling of forest fires, but this time dealing with the fires at the scale of a valley. It'll require to build new stereo vision devices, and also to use near infra-red cameras, and in the end to exploit AI. We can consider that this will be our work for the next 5 years.