Research On Key Technologies Of Early Smoke Feature Extraction And Flame Detection Based On Video Analysis In Traffic Scene

Traffic fires occur frequently,which may easily lead to mass casualties and injuries,causing widespread concern of the media and public opinion,resulting in adverse social impact.How to prevent and reduce traffic fires has become a public safety issue needs to be solved urgently.Traffic fires have the characteristics of strong sudden,difficult evacuation and high frequency of induced derivative accidents.Fire response should be based on “prevention and combines prevention and elimination”.In the aspect of prevention,fire detection devices are used to monitor the fire situation and improve early warning capabilities of fire.Traditional fire detection devices,such as gass-sensitive,temperature-sensitive,and smoke-sensitive,are difficult to meet the fire detection requirements in traffic environments,for reasons that they are susceptible to environmental influences.Fire detection based on video analysis has become a research hotspot.At present,some research results have been accumulated,but there are still quite a few problems,such as high false alarm rate and weak adaptability of traffic scene.In view of the numerous traffic scene,such as highways,parking lots,etc.,it is difficult to traverse the research.In order to facilitate the research and ensure experimental safety,the paper initially selects the urban auxiliary road as the research scene,accomplish the segmentation of suspected smoke and flame areas,then analyze smoke and flame characteristics by multi-feature fusion,for exploring the common characteristics of fire target recognition and early warning,to realize fire monitoring in traffic scene.The thesis aims to identify traffic fire targets by video analysis method,and lay a research foundation for shortening traffic fire response time,realizing early warning of traffic fires and ensuring traffic safety.The main research contents are as follows:(1)Through the experiment comparison and analysis of common fire target detection model,based on ViBe algorithm,combined with two frame difference method and mixed Gaussian model to improve the suspected smoke extraction algorithm,which tacking into account the real-time performance,suspected smoke extraction accuracy and robustness.(2)By using the regional growth method,combined with the smoke LBP feature to segment the traffic smoke area;based on this,extract the dynamic and static characteristics of traffic smoke such as smoke irregularly,HIS color space and smoke entropy characteristics.(3)Establish the candidate area of the suspected flame target,extract the spatial distribution characteristics and stroboscopic characteristics of flame color,design the flame recognition algorithm,and provide the basis for the comprehensive judgment of fire.(4)Integrate the smoke and flame identification algorithm in traffic scenes,design the traffic fire monitoring software based on video fire recognition based on MFC framework and OpenCV library.Through testing,the software realizes the functions of early traffic fire monitoring and warning,visualization of monitoring status and real-time reporting of fire source information.It has better anti-jamming ability,real-time performance and robustness to non-smoke targets.