Research On The Algorithm For Fire Smoke Detection In Video Using Dark Channel Prior

Fire detection based on video-image performs well in respond rate, detection range and capacity of resisting disturbance. So it is more efficient than traditional methods in fire detection for large indoor and outdoor space. Smoke appears earlier and spread further than fire flame in the process of fire. Hence, smoke detection plays a key role in early warning of fire and saving time for people evacuation and rescuing belongings.Smoke detection becomes more and more appealing because of its important application in fire protection. In this paper, we improve the traditional frame difference method to extract suspected smoke areas and propose a new method for smoke detection using dark channel prior based on study and analysis of the existing smoke detection methods. The dark channel of smoke images appears bright, while the dark channel of smoke-free images is dark really. We also suggest some more universal features, such as the changing irregularities of the contour of smoke, the motion direction and semi-transparent. In order to integrate these features reasonably and gain a low generalization error rate, we propose a support vector machine to learn these features and classify smoke images. The feature set and the classifier can be used in various smoke cases contrary to the limited applications of other methods.Experimental results on different scenes show that the algorithm is reliable and effective. It can detect the smoke but also the density of smoke compared with other video smoke methods.