Study On Tunnel Flame Detection Technology Based On Video

With the rapid development of tunnel construction, the tunnel fire accidents occurfrequently, due to the closed structure of the tunnel, the consequences of the fire is moreserious. The traditional fire detection technology because of various interference factors, hascertain limitations, the fire detection based on video technology has attracted more and moreattention, which compared with the traditional fire detection technology has the advantages oflow cost, good real-time, high accuracy, strong anti-interference etc.. This paper mainlystudies the tunnel flame detection technology based on video.This paper firstly proposes an improved statistic histogram method to extract thebackground image, the algorithm selects the higher credibility peak point as backgroundaccording to the energy value around the peak point in the statistical histogram. Meanwhile,in order to improve the reliability of the background, this paper uses a new backgroundupdating algorithm based on block stability, the algorithm through the block matchingoperation to judge the block’s stability, for the block reaches a certain stability degree usingthe weighted average method to update the background.After obtaining reliable background, the paper firstly adopts block binarization methodbased on background subtraction to detect the moving targets, and then respectively usesmorphological method and connected component labeling method for filtering and labelingprocessing of the detected results. In order to eliminate some interference regions, improvethe efficiency of the post processing, according to the characteristics of the flame studied, thispaper presents the flame regions extraction algorithm by combining brightness and stability,using the high brightness characteristic and the relatively stable characteristic of flame toextract the target areas.In order to improve the recognition rate of the system, in this paper, for the flame regionextracted, its area diffusion characteristic and flicker characteristic are analyzed according tothe overall growth rate of area and the change of area in a short period of time respectively.The algorithm designed in this paper has been tested in different traffic scenarios, theexperimental results show that the algorithm can effectively detect the fire incident, has goodreliability and practicability, at the same time, it and can meet the real-time requirement of thesystem.