Research On Flame Detection Technology Of Highway Tunnel Based On Video

Highway tunnel fire accident is one of the most harmful hazards in the tunnel safety accident.It is of great significance to quickly and accurately detect the early fire in the tunnel.The traditional fire detection technology is affected by its inherent characteristics and environmental factors.It has a lot of defects and deficiencies in the detection of early fire in highway tunnel.Video fire detection technology with its high accuracy,real-time strong,strong anti-interference,etc.,suitable for use in highway tunnel environment.Based on the road tunnel environment,this paper studies the movement characteristics of flame,studies the distribution of flame pixels in color space,analyzes the feature extraction method of flame,and proposes a video-based road tunnel flame detection method based on data fusion technology.Firstly,the paper introduces the video fire detection system in the tunnel.For tunnel early fire and flame characteristics are analyzed in detail,and the combination of static and dynamic characteristics,choose to characterize the early fire flame tunnel;and analyzes the relevant theories of the video flame detection technology.Lay the good foundation for the research of flame detection algorithm.Secondly,in order to obtain the flame candidate region with high high accuracy and low interference in the fire image,the image preprocessing method is studied.On this basis,the flame candidate area is determined by moving target detection and color division.For this reason,firstly,through the study of the motion detection algorithm,the hybrid Gaussian model is used to establish the reliable background,and the background difference method is used to detect the moving target.Then,the distribution of flame pixels in the color space is studied in detail.In the RGB space and YCbCr space,the segmentation algorithm of the flame region in the image is proposed to determine the suspected flame color region with high accuracy.Finally,the moving target area and the suspected flame color area are matched to determine the flame candidate area.And the morphological processing,the connected domain markers and the boundary extraction are carried out to prepare the feature extraction.Thirdly,the flame characteristics are extracted,and the color distance,area change rate,circularity,flicker characteristic and texture feature of the flame candidate area are extracted respectively.Then,the support vector machine model is trained by using these features,and the model Video in the identification of the flame,The results show that the model is very accurate.Finally,this paper constructs a video-based highway tunnel flame detection experiment system.Based on MATLAB R2012 a,a software platform is developed to test the flame video algorithm with various scenes.The results show that the proposed flame detection algorithm can accurately detect the flame in the video,effectively filter out the interference of the sports car and various lights,with high accuracy and anti-jamming,can be applied to highway tunnels Fire detection.