Research On Video Flame Detection Method Based On Multi-feature Union

From the use of fire by humans to the present,fire has always been a hazard to human life and property.With the development of society,people's material living standards have been continuously improved,and the damage caused by fire has also gradually increased.In order to prevent and avoid the occurrence of fire,people have invested a lot of energy,human resources,material resources and financial resources.Traditional flame detection methods are based on sensors that identify flames by detecting flame characteristics such as temperature,smoke,and light,but this method is susceptible to environmental interference.With the increasing maturity of digital image technology and the widespread use of video surveillance,video flame detection technology has been developed and developed.The technology analyzes the information in the surveillance video to automatically identify the flame and alarm,which has the advantages of simple deployment,fast response,high accuracy and so on.This paper first introduces the background and significance of the flame detection technology research,and briefly summarizes the advantages and disadvantages of the sensor-based flame detection technology and the video-based flame detection technology,and focuses on the development status of the video flame detection technology at home and abroad,and the research results..The flame motion and color information analysis methods are studied and designed.The flame multi-dimensional feature extraction method is designed and the multi-feature combined video flame detection is realized using Naive Bayes.The main research work of this article is as follows:In the process of analyzing the flame detection technology of video surveillance,a motion detection method based on neighborhood energy background subtraction is proposed to improve the correlation between adjacent pixels and reduce the influence of noise.Based on the analysis of the three primary colors of light and the flame color of the fire image,a color model with red-yellow-blue(RYB)is proposed,which can segment the suspected flame region effectively and accurately.Flicker is the obvious visual characteristic of the flame.Based on the former research methods,and by analyzing the spatial scintillation energy intensity of a large number of flame and non-flame pixels,the flicker characteristic of the flame combustion process is combined with the energy information of the space,then according to the difference of the spatial scintillation energy intensity of the pixels,can obtain a good discrimination effect of the flame and non-flame pixels.A temporal-spatial high order derivative local binary pattern based on circular shift sub-uniform is designed to extract the texture features from the suspected flame region.Finally,the extracted the motion features,color features,spatial scintillation energy features and TDLBPC texture features are analyzed synthetically to realize the flame detection of the video.The experimental results show that the proposed method is more robust,the motion detection and flame color detection are more reliable,which reduces the influence of non-flame moving objects and having similar color with flame objects.