Research And Implementation Of Intellective Fire Detection System Based On Infrared Images

Fire is a common cause of serious natural disasters. Its early real-time monitoring is the focus of research in the field of fire control. Existing fire sensors, such as smoke, temperature, and photosensitive detectors respectively use the smoke, temperature, and optical properties of the flame for fire detection. For their limitations of detecting principles it’s very difficult to find early hidden trouble. The infrared image is the extension of human vision. Infrared-image fire detection technology can obtain fire early warning and early protection through the analysis of the characteristics of early fire images, which is of practical significance to prevent or reduce the occurrence of fire.In this paper, we mainly make use of the theory of infrared radiation, do in-depth study of the characteristics of fire and achieve the temperature changes, area changes, edge jitter and shape irregularity and other characteristics from the collection of infrared images in a fire based on image recognition and processing method. We carry on corresponding algorithm researches and give the concrete mathematics model to build an infrared image intelligent fire detection system based on OpenCV (open source computer vision library).In this paper, the main algorithms include temperature algorithm, area algorithm, shape algorithm and so on. Temperature algorithm mainly bases on the property that when the temperature of the object is higher,its radiation energy is bigger. Accordingly, the image pixel value is higher. By setting the appropriate threshold, it can filter low temperature background information;Area algorithm mainly bases on the property that early fire area will be increasing, which is the notable feature distinguished from other high temperature objects. As the fire increases, the flame continuously strengthens. At this stage, the fire area present continuous, extended trend. It’s very useful for detector to detect early fire through initial growth rate of radiation area;Shape algorithm mainly bases on the property that the flame shape is changing all the time. The shape changes of fire flame reflect the space distributions of flame. According to the irregularity of the fire flame shape and regularity of most of the interference sources (such as a flashlight, incandescent lamp), we regard circular degree as one of the criteria of fire.The system integrates real-time monitoring, image recognition, parameter adjustment, video recording, video capture, logging, burning objects recognition and alarm functions, support the field monitoring and video evidence (fire black box), and can realize remote network alarm and can be linked to various types of fire equipments. The burning objects recognition has important sense.It can reveal the causes and responsibility of fire; It can inform the firefighters the properties of fire combustion. For example, oil should not be extinguished by water gun.