论文部分内容阅读
以洞庭湖流域为研究区,对大范围湿地信息遥感提取方法进行了研究。先基于时间序列MODIS EVI及物候特征参数,通过J-M(Jeffries-Matusita distance)距离分析,构建了MODIS(250 m)最佳时序组合分类数据;其次,通过Johnson指数确定了最佳分割尺度,采用面向对象的遥感分类方法(Random tree分类器)提取了洞庭湖流域的湿地信息,并验证该方法的适用性。研究结果表明,基于时序数据与面向对象的Random tree分类的总体精度和Kappa系数分别为78.84%和0.71,较之基于像元的相同算法的总体分类精度和Kappa系数分别提高了5.79%和0.04。同时,基于面向对象方法的湿地整体的用户精度与生产者精度较基于像元方法分别提高了4.56%和6.21%,可有效提高大区域湿地信息提取的精度。
Taking Dongting Lake basin as the research area, the remote sensing extraction method of large-scale wetland information was studied. Firstly, based on the time series MODIS EVI and phenological parameters, the optimal combination of MODIS (250 m) time-series data was constructed by JM (Jeffries-Matusita distance) distance analysis. Secondly, the optimal segmentation scale was determined by Johnson index, The remote sensing classification method of object (Random tree classifier) extracts the wetland information of Dongting Lake basin and verifies the applicability of this method. The results show that the overall accuracy and Kappa coefficient of random tree classification based on temporal data and object-oriented are 78.84% and 0.71 respectively, which are 5.79% and 0.04 higher than the same pixel-based algorithm respectively. At the same time, user accuracy and producer accuracy of wetland based on object-oriented approach are respectively increased by 4.56% and 6.21% compared with that based on pixel method, which can effectively improve the accuracy of wetland information extraction in large area.