Hyperspectral Image-Based Night-Time Fire Detection Using NKNBD

Abstract

Fire detection on roads at night has many limitations in the fire detection technique using ordinary RGB cameras. It has the problem of misinterpretation of flame due to various lighting sources such as vehicle lighting (LED, halogen) and street lamp (LED, fluorescent lamp). Hyperspectral cameras have hundreds of bands that can distinguish light sources based on the spectra of the light source. Therefore, it is possible to distinguish between the light of the flame, the vehicle light on the road, and the light of the roadside tree. In this paper, a method to simulate a fire in an environment with vehicle lighting and streetlight and detect a fire using NKNBD (Normalized K and NIR Band Difference) in hyperspectral camera is discussed.