Low-illumination Indoor Smoke Detection Method Based On Video

Fire smoke detection technology based on video is a popular research direction in firesafety field over the past decades. Many researchers have devoted to the research of thisaspect, and put forward some effective detection methods. But these methods were aimedat well light. Low-illumination indoor smoke detection method based on the video was afield of less research results. Therefore, this paper proposes a low-illumination indoorsmoke detection method based on video, using camera with infrared source.Because of insufficient illumination of infrared source and the camera’s imperfection,the captured video image under low illumination contains lots of random noise. Thealgorithm firstly reduces noise of gray image by using median filter. Secondly, thesuspicious area is extracted using the combination method of background difference andframe difference based on the threshold segmentation of OTSU. Finally, the area ofremoval noise by morphological filter is analyzed using smoke characteristics.Compared with image at well light, low-illumination image almost lost all of thecolor information and most energy information, and the smoke also lost transparency. Thealgorithm analyze the suspicious area using three smoke characteristics, diffusivity,irregularity, texture, that low-illumination image of smoke can still reserve. Firstly,method of diffusivity analysis is obtained by the fact that the area of smoke is expandingat the beginning of the fire. Secondly, method of irregularity analysis is achieved throughcalculating the circularity of smoke’s contour. Finally, method of texture analysis based onGLCM is obtained, because graphic element of smoke has high definition and its gray iswell-distributed. Finally, fire alarm will arise if the suspicious area meets the threecriterions.The smoke detection algorithm was tested in the real-time circumstances of differentsmoke concentration, shoot distance, light intensity and disruptors. Experimental resultsshow that the smoke detection algorithm performances well on smoke detection while thedistance between the camera and smoke source is less than5.5meters.