A Video Smoke Detection Algorithm Based On Multi-Feature Fusion

The early fire is often accompanied by the emergence of smoke,making it so significant for the timely accurate detection of smoke to improve the accuracy of fire early warning and protect the safety of people's lives and properties.The traditional smoke detection methods,generally by means of the smoke,heat and other sensors,are conducted according to the variation of some physical quantities produced by the smoke,such as the particles,the temperature and so on.However,such kind of detection technology may not alarm effectively until a certain high concentration of smoke,leading to the missing of the best rescue time.In contrast,the video-based fire smoke analysis technology can give an effective alarm immediately as long as the smoke occurs.To overcome the deficiencies of traditional fire smoke detection technology,this paper proposed a new smoke detection algorithm based on features according to the characteristics of smoke movements,so that it can improve the detection rate of smoke detection algorithms.In particular,the algorithm extracts the moving regions by combining the Gaussian Mixture Model(GMM)for background modeling with the mean of background subtraction division method at first.Then use the smoke color and fuzzy characteristics for testing,if meet,into the next phase of testing;Otherwise returns on the step.Later,by dividing the motion area into three parts,including the upper,middle and lower part,the algorithm extracts the optical flow vector feature and edge orientation histograms from each part.Considering the continuous relevance of smoke movement in the time domain,the algorithm extracts the feature vectors of smoke every three adjacent frames to enhance the robustness.Finally,the training and detection of smoke are implemented by using Adaboost.In this paper,many test videos were recorded on different occasions to test the accuracy of the algorithm.A high detection rate above 94% is obtained on the video test set.The experimental results show that the proposed algorithm can better adapt to the complex environmental conditions in practical applications compared to other existing algorithms.