摘要
EOS—MOD IS数据在森林火情监测中的应用研究日益受到世界各国的重视。为了获得适用于中国不同地区森林火情监测的成熟技术,很有必要对现有MOD IS数据林火监测理论算法进行验证分析,探讨其在中国不同地域和季节中使用时的通用性。为此,利用中国境内9起森林火灾事件对MOD IS数据火点识别的理论算法进行验证分析。结果显示9起森林火灾有8起被有效检测到,1起森林火情被遗漏。通过对9起森林火点及其邻近像元的统计分析,发现如下两个重要规则:利用火点亮温偏离统计均值3倍标准差的关系来确定阈值,可以避免火点的遗漏;林火点在CH 21和CH 22上的亮温值一般有CH 21-CH 22<20 K,而噪声点在两个波段上的差异却比较大。用以上规则改进的MOD IS林火热点识别算法可以检测出用来验证的全部9起林火事件,从而证明了改进算法的有效性和通用性。
It has been attached importance to monitoring forest fire based on EOS-MODIS data in the world. How far is the universality of the arithmetic to identify forest fire based on MODIS Data. There still exists much demand for validate the arithmetic. Nine forest fires occurred in China is used to validate arithmetic of identifying fire based on MODIS data in this paper. The result shows that the arithmetic is universal mostly. Eight out of nine fires can be detected through the arithmetic based on MODIS data for identifying forest fire . Just low temperature hot spot could be missed sometimes when zone or season changes. Threshold value should be defined through relation among light temperature of fire and mean and standard variance of fire neighborhood ,which could avoid missing fire spot. There are two important rules when identifying forest fire by MODIS data as follow. ① Light temperature of forest fire is three times standard deviation above the mean from fire pixels and its neighborhood ;②Light temperature of forest fire is near on the band 21 and band 22 ,which is usually characteristic of CH21--CH22〈20 K instead of noise that. Therefore, the rule CH21--CH22〈~20 K can be used for distinguish fire from noise. Additional, it is effective to distinguish forest fire from other fires spot by using forest bound data which extracted from MODIS data without cloud before fire date. Finally suggest for improving arithmetic of identifying forest fire based on MODIS data is presented according knowledge acquired.
出处
《遥感技术与应用》
CSCD
2006年第3期206-211,共6页
Remote Sensing Technology and Application
基金
国家重大基础研究项目(973)前期研究专项资助(2003CCA02100)
福建省科技三项基金(K04016)资助
