Fire alarm system based-on video processing

Abstract

Fire is one of the disasters that often occur in everyday life and causing losses both in terms of material and non-material. There has been much research done to build a field alarm system. One way to build a fire alarm system is base on fire detection on video data; this is done with digital image processing techniques and machine vision. Fire detection systems based-on video processing consist of four main stages, namely: motion detection area, detection area of the color of fire, temporal analysis of probabilistic calculations or applying wavelet analysis, spatial analysis using energy calculations on an area. In this research, we focused on selection method in temporal analysis phase in order to build a more responsive fire alarm system. Fire alarm will be activated when the system detects the occurrence of wild-fire in a certain position and time interval. Alarm will be activated if the occurrence of wild-fire in a certain position and time interval surpasses the threshold. The occurrence counters calculate by comparing the historical location of occurrence centorids with the centroids of wild-fire that being detected. Experimental results show that color probabilities on temporal analysis provides more accurate results in the detection of fire, but it is slower than wavelet analysis to detect the first occurrence of fire. We recommended the use of wavelet analysis in temporal analysis stage in order to produce a responsive system.