Research On Fire Detection Rules Based On Flame Recognition

The process of Fire Alarm (FA) can be divided into 3 parts, signal detection, signal analysis and signal transmission. The purpose of signal detection and signal analysis is to make sure whether there exists destructive combustion or latent destructive combustion. Combustion will produce new materials and energy, such as smoke, CO, CO2, heat, light (flames) and noise etc. The object of signal detection is the materials and energy produced by combustion.Therefore, in terms of signal detection Fire Alarm Systems (FAS) can be classified into smoke FAS, temperature FAS, image FAS, and gas FAS etc. In fact whether combustion can produce smoke or how much smoke it can produce depends on the completion of the combustion. So smoke is not an essential feature of combustion. And what gas combustion can produce depends on what material is burning. Materials that don't contain carbon will not produce CO when burning. Generally combustion is an exothermic process. So producing heat can be taken as an essential feature of combustion. However, the speed of heat transmission in air is slow. And the detection range of temperature sensors is limited. In order to monitor a certain space completely many sensors are needed. The cost is too high.Image FAS developed in recent ten years overcomes the limitation of detection range. And the detection object, flame, is a common feature of combustion. Usually rules for flame detection depend on features of color, luminance, shape and their temporal and spatial development. According to the shortcomings and faults of traditional color and luminance model of flames this dissertation presents the spatial distribution features of flame luminance and the relation between Red Saturation (Rs) and Luminance (Y).Based on the Rs-Y detection rule, this dissertation makes a further recognition by flame area detection rule and flame sharp angle detection rule. For stable fire the flame area doesn't change much with time. For unstable fire the flame area increases with time. We set a threshold for the increasing rate of mean area of the consecutive flame images to distinguish stable fire and unstable fire. However, when stable fire is moving towards camera the area of their flame images also increases. In order to exclude this we take the detection rule based on flame sharp angle. The number of sharp angles of unstable flame changes more evidently than that of stable flames. The experiment results show it has a good performance of accuracy and anti-interference.The flexibility of this design is also on that it doesn't need a monitoring center. The alarm receivers are 3G cell phones. So it is quite appropriate for homes. And you don't need to press any keys on the fire alarm machine to control it and set its parameters. All these functions can be completed by short messages. The abrasion is slight and the alarm machine can be controlled in remote places.