摘要
Landsat卫星遥感数据具有分辨率较高,数据积累时间长的特点,在探测地表覆盖变化和地物分类中得到广泛应用。首先,对获取的Landsat TM/ETM＋时间序列数据进行了定量化处理,获取了三江平原七台河市1989--2012年时间序列Landsat地表反射率图像。其次,设计了林地指数和湿地指数,提取了三江平原七台河区域地物光谱和时序特征,同时设计构建了地表覆盖分类和植被地表类型变化探测的决策树算法,实现了1989--2012年七台河区域的植被地表覆盖变化的动态监测,提取了森林覆盖变化的空间分布与变化时间。最后,对七台河区域地表覆盖与植被地表类型变化进行了精度检验,分类总体精度达到90.04%,Kappa系数达0.88。研究结果表明：基于定量化的Landsat时间序列数据的分类算法能克服单时相影像分类的缺陷,实现区域地物自动分类和地表覆盖变化的动态监测。
Landsat satellite data is widely used in monitoring land cover change and land cover classification by its medium resolution and long time-series records.In this paper,twenty Landsat TM/ETM＋images in Qitaihe district in Sanjiang Plain were collected,and quantitatively processed for time-series ground surface reflectance stacks from 1989 to 2012.Then,the forest index and wetland index were designed,and the spectral characteristics and their time-series variation features of different land covers were extracted from these time-series reflectance stacks.Thirdly,a decision tree-algorithm was designed to classify different land covers and detect the temporal change of vegetation land-types from 1989 to 2012.Finally,the classification result was validated by the ground survey data,with an overall precision of 90.04%,and a Kappa coefficient of 0.88.The result proved the potential of time-series Landsat images for land-cover and land-use change.
出处
《遥感技术与应用》
CSCD
北大核心
2015年第5期959-968,共10页
Remote Sensing Technology and Application
基金
中国科学院对外合作重点项目(GJH21123)
国家自然科学基金项目(4122208)
