Résumé:
The present thesis traits the problem of detecting forest fire from high resolution satellite image. Our methods are applied to a single landsat 7 satellite image of a fire occurring in Californian zone in July 2006, USA. A forest fire is an outbreak that propagates a wooded extent. It can be natural (due to a lightning or a volcanic eruption) or human (caused by mankind). Worldwide, the impact of forest fires on numerous aspects of our daily life is becoming increasingly noticeable, particularly over the products fluxes we depend on, the health and reliability of the communities we live in, and especially on the state and maintenance of natural ecosystems. Many methods have proved the effectiveness of detecting forest fire. The eccentricity of this work lies in the detection of forest fire thanks to thermal infra-red satellite image using the belonging probability inverse image. Finally, to clarify the efficiency our suggested algorithm, we have tested it over different satellite image. The obtained results which were compared to those of Florent Lafarge are more accurate.
