Research On Smoke Detection Method Based On Image Analysis

The traditional fire detectors can detect smoke in the fire early,this is a contact detection type method,it has a good detection effect in room.But for the outdoor,this traditional detection method based detectors has a bad effect for smoke detection.In order to prevent fire and fire extinguishing in a large space environment,fires can be detected in real time and effectively,this article introduces a smoke detection technology based on video image analysis with better adaptability and versatility.By analyzing the real-time image of the monitored area,the analysis based on color,texture,shape,flaring,and flicker frequency is used to determine whether smoke is present,to determine whether there is a fire or potential fire,and to provide fire warning.Reduce various losses caused by fire.In order to detect effective smoke areas from the collected video surveillance images,the following three aspects were studied:1.The existing image acquisition system are discussed in this paper,the general image acquisition system mainly includes two types about field of view.The one has small view field but has a good imaging effect,the other has a large field of view which called panoramic image acquisition system.Its application in outdoor environment is compared for two image acquisition system.For the monitoring of large seamless coverage is not required,general image acquisition system collected image has small distortion,simple implementation,and better smoke detection effect.2.The preprocessing algorithm was studied which the smoke image collected by general image acquisition system.The algorithms mainly including image denoising,image enhancement,image correction and image sharpening,and the various pretreatment algorithm simulation experiment was performed.The smoke image preprocessing effect were analyzed which collected in different time and different scenarios,then the smoke image pretreatment process is determined.3.The determination and detection of smoke candidate regions from the preprocessed images are studied.Focusing on the characteristics of smoke,three smoke candidate regions determination methods are studied.(1)Research on detection algorithms of smoke candidate regions based on dark-channel priority.Through the dark channel characteristics of the fog image,fog-free images can be obtained.When the two backgrounds are different,the smoke area can be obtained,and the relatively good smoke detection effect can be obtained;(2)Smoke characteristics based on color features In the algorithm study,because the main constituents of the smoke are water vapor and carbides,which have a high luminance value and a low colorvalue at the initial moment of the fire,smoke is detected using the color characteristic of smoke;(3)based on convex grouping The technology of smoke detection method research,convex structure is a unique convex attribute structure of each goal,is an optimization of mathematical methods,because the direct analysis of the target convex properties,with fast,stable target detection characteristics.4.The detected smoke candidate area contains smoke and also contains some smoke-like targets.In order to accurately detect smoke targets from these candidate areas,this paper applies an optimization algorithm to the features of the detected smoke candidate areas.It makes decision decisions,finds the best smoke area and outputs warnings,etc.,to provide a basis for further applications.Finally,the experimental results show that the two steps of smoke detection method using image analysis can better detection the smoke area from the field in the different environment and time to collect images,the research results will provide certain theory and practice of fire smoke detection support,it has a good practical application value.