Reserach Of Fire Detection Method In Highway Tunnel Based On Video

In the operations management of road tunnel,tunnel fire is the the most serious disaster,which will cause great damage to passengers,tunnel and other facilities.If the fire is detected and alarmed in the initial stage,it will help passengers’ evacuation,prevent the spread of fire and reduce the losses caused by the fire.Therefore,it is of great significance to study early fire detection method.Taking road tunnel as the application background,the thesis study the fire classification and detection algorithm using the T-S model fuzzy neural network(Fuzzy Neural Network,FNN)of intergrating the features of flame and smoke.The main contents are as follows:(1)Moving target detection.Using background subtraction method of Gaussian mixture model to extract the moving target,and adopting algorithms such as dilation,median filtering and pin-hole filling to process the images of the moving area.(2)Suspected area division.Combining fire flame of RGB and HSI color model is used to segment candidate flame area;and the component color statistic model of the fire smoke of RGB and I is used to segment the suspected smoke area.(3)Extraction of the flame’s features.Researching the layered characteristics of the color area of the flame,and giving the stratification feature model of color area of the flame;Researching and realizing the features of the first and second order moment ratio in the flame’s Local Binary Patterns(LBP),and the beat frequency features of the flame.(4)Extraction of the smoke’s features.Researching and giving smoke texture feature of combing LBP and the gray level cooccurrence matrices(Gray Level Cooccurrence Matrices,GLCM);Studying and.implementing smoke diffusion stability characteristics and its extraction methods.(5)Fire detection algorithm design.The paper gives a fire detection algorithm based on T-S model FNN.T-S FNN fire detection network structure of the fusion flames and smoke’s multi-feature is established.T-S FNN model is trained offline by using flame or smoke training samples,and made an offline simulation test.The T-S FNN model of a higher classification accuracy is chosen as fire detection model in the paper.The paper respectively conducts offline simulations and online tests,based on the T-S FNN flame,based on T-S FNN smoke,and based on BP neural network fire classifiers,results of comparative experiments showing the feasibility and effectiveness of the T-S FNN detection algorithm.Fire detection algorithm in the paper fuses image characteristics of the fire flame and smoke,and through detecting videos such as the highway tunnel simulation fire and driving interference under different lighting conditions,the results show that the algorithm can effectively identify the fire,and rule out the effect of the tunnel distractors.Therefore,the detection algorithm mentioned in the paper has a good practical application value.