摘要
以2012年4月20日在蒙古国发生的特大草原火为例,用HJ-1B卫星CCD影像数据作为数据源,采用支持向量机分类方法提取火烧迹地,再用目视解译数据进行精度验证,总精度达到82.71%,Kappa系数为0.666 8,达到高度一致水平。对灾前的植被指数进行分析后,发现NDVI值小于0.1的比例为0.46%,大于0.1的比例为99.54%,因此能够判定当NDVI值大于0.1时,发生和蔓延草原火的概率较高。
Taking the large prairie fire occurred in Mongolia on April 20,2012 as an example,the burned traces are extracted by using HJ-1B satellite data as the data source and classification method by support vector machine and the accuracies are verified by visual interpretation data,and the total accuracy reaches quite a high level as 82.71% with kappa coefficient of 0.666 8.Normalized Difference Vegetation Index（NDVI） before the disaster are analyzed to be with a ratio of 0.46% less than 0.1 and 99.54% greater than 0.1,which indicates that when the NDVI value is greater than 0.1,the probability of occurrence and spread of grassland fire are high.
出处
《灾害学》
CSCD
北大核心
2013年第1期32-35,共4页
Journal of Catastrophology
基金
公益性行业(农业)科研专项经费(200903041)
国家自然科学基金(41071326)
关键词
HJ-1B卫星
火烧迹地
支持向量机
草地可燃物
HJ-1B satellite
fire traces
support vector machine
combustibles on the grassland
