Computer Vision Based Smoke Detection Method by Using Colour and Object Tracking

To reduce the damage from fire disaster, demand for automatic detection system by using computer vision technique is increasing. But because of false detections that are caused by various situations, it is hard to use in real. In fire detection area, to overcome this problem, the algorithm using several temporal and spatial information of object is proposed. Colour, brightness, and movement information is used to make relevant smoke detection algorithm. And continuous monitoring for several constraints is used to avoid false detections that are caused by unexpected behaviour of object. In experimental result, total 11 videos that have smoke and the other 2 videos that have no smoke is used for test performance of proposed method. It shows relevant performance for false detection and could detect almost fire in video.