Reduction of false alarm rate in automatic forest fire infrared surveillance systems

Abstract

One of the main problems related to infrared remote sensing of forest fires by automatic systems concerns the rejection of false alarms. The study of the infrared spectral radiance emitted by a biomass fire has been used to define spectral algorithms that permit to separate fires from other sources considered as false alarms. The different behaviour of the medium (3-5 μm) and thermal (8-12 μm) infrared spectral regions has been the basis for the definition of a new "figure" of merit, the so-called Fire Index (FI). An experimental burn has been used to test the ability of FI to identify clearly fires. A second algorithm based on an intraband analysis, the Mid-IR Fire Index (MFI), has been proposed to give complementary information to FI. Evolution of FI and MFI values with distance for different IR sources has been calculated. Finally, an experimental validation of both algorithms has been performed by using two infrared cameras working in the medium and thermal infrared spectral windows.