An Algorithm For Early Warning Of Forest Fire Based On Image Processing

Forest fire prevention is a major event in people's livelihood, and the new fire protection system is undoubtedly great significant. Previously artificial, satellite and other means for detection of forest fires is often a huge waste of human and material resources. Based on this, people want to put forward a new approach to deal with this problem, since then, fire protection systems based on image processing more and more cause for concern.The main problem of Forest Fire prevention is to ask the system can identify images from the fire and smoke, a timely warning and allowing to minimize harm. To identify smoke and fire, the characteristics of smoke and fire will be necessary to fully understand and the interference caused by external environment will be learned, We mainly need to do is to identify fireworks and distinguish between the real smoke and interference. So what characteristics of smoke and fire can take advantage of?First, when considering the color feature, regardless of smoke or fire, they both has its own characteristics in the RGB space. They have different characteristics to take advantage of through the analysis of brightness, color saturation and hue.Then the spread of smoke and flashes of fire can be served as determination conditions of judging fire. In face of a complex external environment, we will encounter many problems:1 A continuously rotating camera can not make the scene be fixed.2 Camera jitter makes it impossible to carry out comparative analysis amongmultiple images on some properties.3 Snow, fog, lights and so on make the identification disruptive.In view of the above, we solve this problem in two main directions. To put it simple, they are:fire and smoke detection on the space domain and time domain. In the testing process, we also analyze in both directions. Analysis on the spatial domain.The analysis of the single-frame image is main emphasis. At this point the algorithm is less affected by the lens rotation and jittering, but the characteristics of smoke and fire can be used is little too.Determination conditions of fire: R>G>B (1) R>Rmean (2) 0<=H<=60,0.2<=S<=1,100<=v<=255 (3) Of which:Rmean= 1/KΣR (xi,yi).Determination conditions of smokeFirst, we decide the RGB values of the standard smoke through the k-means clustering algorithm and experimental result analysis and the standards to determine whether a particular pixel is smoke pixel:||P-C||2