Research On Fire Detection Technology Based On Video Image In Complex Background

Once fire happens,it will threaten people' s lives and property seri ously and cause huge losses.The rapid developm ent of science an d technology has prom oted the technological innovation in many fields.The technology of fi re detection based on video and im age has aroused public attention.Com pared with the traditional fire detection technology,the fire detection technology based on video i mage not only has the advantages of high detection efficiency,large monitoring scene and no environment restriction,but also can further satisfy people's intelligence and information demand for fire detection system through the application of artificial intelligence technology.On the basis of the ex isting video image fire detection technology,this paper introduces the theory of gray level co-occurrence m atrix,wavelet transform and depth learning to fire detection,and has carried on an in-depth study.The following is the main work of this paper.(1)Propose an improved flame segmentation algorithmAn improved flame region segmentation algorithm based on HSI space color correlation,two-dimensional Otsu algorithm and seed region growing algorithm is proposed.The algorithm can automatically segment the suspected flame region,which not only overcom es the influence of high brightness interferon in the background of the fire image,but also avoids the edge blurred when the edge seed points are growing in the region.(2)Propose a fire detection algorithm based on gray scale co-occurrence m atrix and wavelet transformThe static and dynamic features are extracted from the two aspects of the texture details and the edge changes of the flame,and then the feature vectors are formed.The support vector machine is used as classifier based on these feature vectors and the results of the fire detection are given.Through the open database verification,the algorithm performs well,and has a high recognition rate for fire images under complex background.(3)Proposes a fire detection algorithm based on convolution neural networkApplying deep learning technology to the field of fire detec tion,a fire detection algorithm based on convolution neural network is proposed.Through the input of the original video image,the deep level of fire featur e information can be ex tracted to av oid the shortcomings of describing the fire image by artificial selection of the corresponding features.Compared with the conventional fire detection algorithm,the fire detection algorithm based on convolution neural network performs better.