Research On The Method Of Fire Detection Based On Video Image

Conventional fire detection system, which is based on the smoke sensor or infrared sensor, is relatively successful at present. It is also widely used in the market. However, for large warehouses or other large space, such as railway stations, airports, forests and so on, it impossible to make fire alarm timely and accurately, because the smoke and temperature information are not obtained quickly. In order to solve this problem, it is necessary to study fire detection technology based on video.In this paper, the static and dynamic characteristics of the flame are analyzed to establish the flame’s color model. The suspected fire area is extracted based on the model. The objects of resembling the flame color are removed based on moving target detection. The dynamic characteristics for the suspected fire area, such as flicker feature, area feature, are extracted to determine whether the area causes a fire.In this paper, the characteristics of smoke are analyzed. They include the color of smoke, the area growth, the fire smoke area of irregular shapes and the blurred background which the smoke causes. The smoke image is analyzed statistically to establish smoke color model. Because the smoke causes blurred background, the high-frequency energy attenuation rate of the image is used to describe fire smoke and the smoke area is detected effectively.The fire detection method based the combination of flame and smoke is used to design and realize the fire detection systems in this paper. Experimental results show that the proposed method improves the accuracy and real-time of fire detection.