Forest Fire Point Extraction Research Based On MODIS Data

Forest resources is a very important kind of natural resources, forest fire has become the major cause of the loss of forest resources, so the detection of forest fires has been more attention. The remote technology gives the support of the forest fire monitoring and disaster assessment. The EOS-MODIS data is used in this article's experiments. In the practical application of has a well application prospects.According to the characteristics and attribute to composed multi-temporal MODIS images of pretreatment, including the Bowtie remove treatment, image geometric correction, atmosphere correction and radiation calibration. Next the c orrection data removes the cloud pixels and water pixels. The algorithm of fire detecting in this paper is derived from the method of fixed threshold of NOA A/AVHRR data. The traditional algorithm needs a serial experiment to confirm the threshold. However, this paper determines the threshold using feature para meters of fire pixels and background pixels on condition that it reserves the m ain idea of traditional algorithm. The new algorithm can reduce the computatio n amounts and improve the efficiency. The new algorithm in this paper doesn't use accurate threshold, which make it more adaptive.In this paper, the procedure of fire burning is divided into two parts, fi re spot burning and smolder burning. According to the brightness temperatur e and burning characteristic, this paper design respective algorithms to detect fire spot and smolder spot. Mean the while, this paper shows the difference of fire pixels and background pixels by using variance between-class. Then, the fire spots and smolder spots can be detected by determining the value of variance between-class. This paper presents an algorithm to detect fire wh ich is more effective to avoiding the erroneous and missing judgment and le ss vulnerable to interference by other factors.