The Study Of An Improved Method For Forest Fire Detection Based On Variance Between-class

Frequent forest fires, seriously damage and threat to the environment and forest resources which human survival in. It is very important.to fighting fires and protecting forest resources. So,the research of forest fire monitoring is urgent. Remote sensing can do lots of works in forest fire monitoring, and it has special potential, as it can monitor wide range of areas,has high temporal and spatial resolution. Especially, The EOS/MODIS sensor is designed for fire monitor and it can give us very great help.Therefore, it is been widely used.The traditional method of fire monitoring identify forest fires with absolute thresholds of some parameters such as brightness temperature. But when it is used in some hot area, some hot background pixel will be mistaken for fire.while it is used in some cold area, it maybe miss the ture forest fire spot. In this article, an improved method using variance between-class and smoke plume mask is described. On the basis of analyzing the characteristic of MODIS data ,author indicates the suitable parameter and wave band used in this improved method At first, the brightness temperature threshold of potential fire pixels was adjusted to be 305K. Based on the variance between-class of TIR channel brightness temperature and a smoke plume detection algorithm, the improved algorithm can separate the hot fire spots from the background and seek out the cool fire spots, respectively, with suitable thresholds of variance between-class. This algorithm has been used in the forest fires happened at Fujian province, Heilongjiang province and outside China. Study shows that detection results with the algorithm are more satisfactory. It is adapted in different environments and can be more accurately detected the high-temperature fire spot and the smoder at low temperature. It creases the ability and accuracy to detect fire spots.In order to evaluate the forest fire damage, Global Environment Monitoring Index is used to identify burnt scar immediately based on identified fire spots.?A case study was carried out in the forest fire occurred in China and outside China. The accurate recognition results not only verify the accuracy of the method of identify burnt scar , but also the accuracy the improved method of fire monitoring.