Research Of Fire Detection Based On The Characteristic Of Video

The fire disaster has a great destructive power for the production and living of people,a serious threat to the people's property security and life safety.Usually the damage which caused by fire disaster in the early stages is not big and easy to be ignored,but its development is rapid and the damage increased rapidly over time,so the sooner you find fire disaster and eliminate it,the damage is smaller.Traditional fire detection methods generally depend on the human inspection,the cost is huge,and can't apply in some special dangerous area;a more common fire monitoring system is based on sensor technology: it combine the sensor of smoke?light and gas with network technology,through the remote monitor and analysis of physical and chemical matter in monitoring places to implement fire disaster monitor.Although this method overcomes the drawback of human inspection,it is also easy affected by environment and difficult to up to the monitoring task on open places.Along with the rapid development of digital image processing technology and computer vision research,the video fire disaster detection technology which combined with digital image processing and computer vision technology with the advantages of low cost,wide applicable scope attracted a large number of experts and scholars both at home and abroad,has become a research hotspot.This article studied and analyzed the research status and the shortcomings of the traditional fire disaster detection technology,based on the analysis of flame images' feature in space and time of video sequence,proposes a fire detection method in video based on digital image processing and computer vision technology.Specific content is as follows:Firstly,an improved ViBe(visual background extractor)algorithm is proposed to extract the foreground of moving flame in video.The brightness matching is introduced into the background model update strategy,in order to solve the original ViBe algorithm cannot distinguish between flame and other moving objects,and when the light changes it easy caused the problem of misuse detection;at the same time,using the interframe difference method to take out the motion region and background as the initialization frame because of its advantage of calculate speed,and then use the the ViBe algorithm for motion foreground accurate extraction,improved the deficiency of the ViBe algorithm that real-time performance decreases with increment of resolution and prone to detect ghost when the selection of initialization of frame is improper.Secondly,deal the extracted foreground area by the flame color feature matching and the flame shape matching.Base on the radio of the number of ordinal raise the match model of the degree of flame angle.And design a detection alarm algorithm based on area growth threshold after the analysis of the growth characteristics of the flame area.Finally,based on Qt and OpenCv,the software of fire monitoring system is developed.The experimental results show that the system can identify the flame information in the video,and has a good prospect in engineering application.