Research Of Identification And Location Of Open Flame Based On Video Image Processing

The fire caused a large number of casualties and property damage each year nationwide,facing to the severe issues of firefighting safety, it is particularly important to increase thereal-time monitoring and detectable efforts. Detecting the fire whether happens or not quicklyand accurately,and extinguishing it in the shortest possible time is the only way to reduce theloss greatly which was leaded by the fire.Currently, there are many methods to detect fire, the traditional method is to use theparticle sensor, which alarms in the scene of the fire through detecting the change of smoke,temperature, gas, light and other parameters. The disadvantages are distance restricted andfalse negatives and false positives serious, furthermore, it is susceptible to be influenced andinterfered by the outside environment and the scope of use and the field of application is notwide enough. In recent years, some scholars have in-depth research in fire detectiontechnology, especially based on image processing technology, which has become a hot spot.But the existing fire detection systems based on image processing technology are oftencomplex and can not get general application in real life. The main reasons is that it extractstoo much features and the algorithm is complicated, which on the basis of preciseness, butextending the detection time, having little practical significance. This paper focuses on openflame detection when the fire broke out, the purpose is to improve the real-time nature,stability and reliability of the fire detection system to achieve the requirement of fast andaccurate.Firstly, on the basis of analyzing the previous research results, we improved thealgorithm of fire identification. Through installing filters in front of the cameras, the visiblelight is filtered out and when fire broke out the light of800nm-1100nm band can high pass.Secondly, use the information fusion technology, fusing only three characteristicinformation of circularity, frequency and sharp corners of the fire image as the criterion of fireidentification. Take the motion of mass center and area change of the fire image as asecondary feature, used to analyze the situation of fire changes. Under the premise ofguarantee of accuracy, the algorithm is simplified and the system improves the detection rate.Thirdly, it can accurately locate the position of fires by using the technology of binocular visual positioning, researches analysis the positioning methods which are based on the linermodel and the camera visual angle. By a large number of experimental data comparison andcomprehensive consideration, we take the position method based on the camera visual angle.Finally, use MATLAB GUI building the integrated operating system of fire automaticidentification and location and it alarms immediately if there is fire, then transport thethree-dimensional coordinates of the location of fire site to water cannons to achieveextinguishing automatic. This system tests well and satisfies the need by multiple field trials.