Research On Smoke Detection Method Of Coal Mine Fire Based On Image Enhancement And Multiple Features

Mine fire is one of the five common disasters in coal mine,which seriously affects the normal production of coal mine and the personal safety of miners.At present,aiming at the special environment of coal mine,there is no mature engineering application of fire smoke detection,and there are some cases such as missing report and false report.Aiming at this problem,this paper discusses the smoke detection algorithm in Coal Mine Based on image features to improve the timeliness and accuracy of fire recognition,which is of great significance to effectively solve the automatic detection and intelligent early warning of mine fire.(1)Coal mine image enhancement.Aiming at the problem that the computer can't detect the smoke accurately due to the uneven illumination and the influence of dust and fog in coal mine,based on the theory of multi-scale Retinex algorithm with color protection(msrcp)and particle swarm optimization algorithm,an automatic enhancement method of coal mine image based on improved msrcp algorithm is proposed to improve the brightness of mine image and highlight the smoke information of mine image.(2)Multi characteristic analysis of mine smoke.Aiming at the moving characteristics of smoke,this paper uses the image difference method based on background tracking to extract the smoke and other moving objects in the mine monitoring video sequence.By analyzing the drift direction of smoke and the background ambiguity,some moving non smoke objects are excluded.Furthermore,the color features of mine smoke are analyzed,and the judgment conditions of mine smoke color are established.Aiming at the unique contour and texture features of mine smoke,rotation invariant LBP features and directional gradient histogram hog features are extracted to describe mine smoke.(3)Smoke feature fusion and judgment.Aiming at the problem that the accuracy of using single feature to detect mine smoke is not enough,a smoke detection method based on feature fusion is proposed,which combines the texture feature and hog feature of suspicious smoke target whose drift direction,background ambiguity and color meet the smoke characteristics.SVM support vector machine is used to determine the final smoke / non smoke of the input feature vector.Through the analysis of a large number of smoke video and non smoke video detection results,this algorithm can effectively restore the face of mine roadway.In the simulation experiment of coal mine smoke video,the accuracy rate of smoke detection reaches 95%,and the accuracy rate of non smoke interference video simulation experiment reaches 99%.The test results show that this algorithm has a certain anti-interference ability,and can meet the requirements of mine fire smoke detection It has a certain practical value to meet the requirements of real-time and accuracy.