Research On Highway Tunnel Smoke Detection Method Based On Video

In the operation and management of highway tunnels,the fire is the most serious type of safety accident.Tunnel fires often cause serious casualties and huge property losses.If it can be detected and alarmed at the beginning of the fire,the loss can be minimized.Therefore,early detection of tunnel fire is so important.Video detection technology is the most effective technology to realize the early detection of highway tunnel fire.In the event of the fire,the smoke is usually earliest detected.In this paper,the smoke in the video is detected by C-SVC recognition method according to the special environment of the tunnel and the various image features of the fusion smoke.The main contents of this paper include the following parts:(1)Moving target detection.This paper analyzes the effects and characteristics of frame difference method,optical flow method,mixed Gaussian background subtraction method and ViBe algorithm in the detection of video smoke.The motion detection algorithm of a ViBe based on GMM is adopted to quickly eliminate ghosting and gets more accurate motion area,then post-processes the image of the motion area.(2)Suspected smoke region segmentation.In this paper,an improved method for segmentation of motion smoke area is proposed,namely the segmentation method of combing RGB,HSI and Lab color space.The experimental results show that the improved segmentation method is effective.(3)a smoke color statistical model segmentation method based on RGB,HSI and Lab color space is proposed to obtain the suspected smoke region.(3)Multi-features smoke detection.This paper mainly studies the characteristics of smoke including LBP Gaussian pyramid,circularity,average gradient and flicker frequency,they can effectively distinguish smoke and non-smoke from the result of experiment.(4)C-SVC smoke recognition model based on SMO optimization.In this paper,the smoke and non-smoke interference samples extracted from the video are composed of the input vector of C-SVC,and the learning model is trained to verify the effectiveness of the proposed algorithm.In this paper,the C-SVC pattern recognition methods are used to detect the smoke and non-smoke interference video.The experimental results show that the method can effectively detect the smoke in the video image and eliminate the false smoke Interference object.