Fire Flame Detection Based On Video

With the development of computer technology, electro-optical imaging technology and image processing technology, video-based fire detection technology has been paid more and more attention .However, the complexity and constantly changing of the fire scene in the real environment; make it very difficult to identify flame. At present, a variety of fire identification method based on video image processing is more effective in a specific environment .When the environment changes, the error of the detection will become great .So far, there is no more mature theories and algorithms. As the temperature difference between inside and outside the flame region, the flame core to the outer flame present different color. Flame colors are actually within certain range, rather than a certain value .In this article, we do plot experiment to the images containing flame. Through the experiment we find the results of the flame segmentation are unsatisfactory in the surrounding environment whose color is closer to the flame. However it is found that I₂ color characteristics has the best segmentation effect in I₁,I₂,I₃ color model. The transition from RGB space to I₁,I₂,I₃ is linear and each component is independent of each other. This paper presents a flame segmentation method based on I₁,I₂,I₃ color model and Otsu arithmetic. This segmentation method can split flame zone to the greatest degree and has good real-time performance in different outdoor environment and different light conditions.According to the motion characteristics of fire flame, first of all, extract the moving area of video sequence through the method of the dynamic three frames. Then take the intersection region to the flame splited out to get the moving area with flame color features. Finally, identify flame according to the flame characteristics of edges, sharp corners, flashing area, to achieve detection purposes. The results showed that this method based on I₁,I₂,I₃ color model and Otsu arithmetic can be very effective to split out the flame zone, and morphological image processing technology enables more accurate segmentation .Then identifying flame according to some salient characteristics of the flame can be effective to the fire detection, which has good practical value.