Fire Detection in Tunnels by means of Image Processing

Abstract

The fire detection is very important to prevent a secondary disaster and to save the assistance for safe driving in the master plan of ITS. At present, the fire detectors, which have a high sensitivity, are used in order to detect the fire in tunnels. It has problems such that the detection area is very narrow and sensor head is fouled by exhaust fumes. Compared with the fire detectors, image sensor is suitable to detect the fire, because it can cover two-dimensional wide space. We had attempted elsewhere the fire detection, which compares two histograms with various shades of gray level, by means of image processing. However, there remained problems about the influence of passing vehicle in the actual outdoors environment. The flame has characteristic such that its state is unstable and repeats growth and decay. A new approach toward the fire detection based on fluctuation of the flame has been described in this paper. Conventional fire detectors could detect only a phenomenon of the fire. However, proposal method could detect not only the flame but also its position. We conducted simulation and field tests to verify the accuracy of this method, and confirmed the effectiveness.