Computer vision based fire alarming system

Fire detection system in the surveillance system monitors the indoor environment and issues alarm as part of the early warning mechanism with ultimate goal to provide an alarm at early stage before the fire become uncontrollable. Conventional fire detection systems suffer from the transparent delay from the fire to the sensor which is looking at a point. The reliability of the fire detection system mainly depends on the positional distribution of the sensors. This paper proposes novel method of fire detection by processing image sequence acquired from a video. The proposed video based fire-detection system uses adaptive background subtraction to detect foreground moving object and then verified by the rule based fire color model to determine whether the detected foreground object is a fire or not. YCbCr color space is used to model the fire pixel classification. In addition to the motion and color the detected fire candidate regions are analyzed in temporal domain to detect the fire flicker. Some Morphological operations are used to enhance the features of detected fire candidate region. All of the above clues are combining to form the fire detection system. The performance of the proposed algorithm is tested on two sets of videos comprising the fire, fire colored object and non-fire. The experimental results show that the proposed system is very successful in detecting fire and /or flames.