Design And Realization Of Fire Detector Software Based On Binocular Vision

In order to avoid fire and reduce the loss of human society caused by fire,people invented a variety of fire detection technology.Traditional fire detection technology mainly detects fire by detecting various characteristics accompanying fire,including temperature,smoke,gas,radiation,particles,sound,and so on.However,the traditional fire detection technology is greatly affected by environmental factors,such as in places with better ventilation,large space,more water vapor or dust,the detector is prone to false alarm or non-alarm.In case of nonalarm,it will bring irreparable loss.In this thesis,a method of fire recognition and location based on binocular vision technology is designed,which can be applied to the situation where traditional technology is not applicable,such as warehouses,factory buildings,and has high fire alarm accuracy.In this thesis,infrared filter technology is used to filter out visible light for fire flame image acquisition.In this way,only objects that can emit infrared light can be imaged on camera CCD,so the influence of ambient light on flame recognition can be greatly reduced.Then,the acquired flame image is grayed,binarized and marked with connected domain(the area less than a certain threshold is not considered as connected domain),and then the suspicious flame region can be obtained.In order to further improve the accuracy of flame detection,this thesis divides the flame suspicious areas into different levels according to their area size,and sets different roundness criteria for different levels.Then according to the maximum area of the connected domain,the detection mode is divided into three types,which are candle mode,lamp mode and general flame mode.According to the area size of the connected region,different dithering frequency ranges are set for the three types of detection modes to construct the fire identification criteria.After detecting the fire according to the criterion,the stereo calibration and stereo matching of the flame images collected by the left and right cameras are carried out by using the parameters of stereo calibration.In this way,the coordinate pairs of the flame pixels in the flame image pair can be obtained.Combining with the parameters of stereo calibration,the spatial coordinate information of the flame can be calculated and the fire location function can be realized.Finally,the binocular vision fire detector software based on this method is implemented on DM816 x platform,and the functions of the fire detector software are tested and checked in the warehouse.200 sets of experimental results show that this method can accomplish the function of fire recognition and location well.The fire detector software can still play an effective role in situations where the traditional fire detection technology is not applicable,such as better ventilation and larger space,and the accuracy of fire alarm can reach more than 98%.