针对桉树人工林在遥感影像上与自然林光谱差异小难以区分的问题,研究选取Quickbird,Landsat 8,数字高程模型(DEM)等多源遥感数据,利用决策树加面向对象的分类方法提取桉树林.在多时相Landsat 8提取兴趣区域(ROI)基础上,利用最优尺度(ESP)工具计算研究区最优分割尺度进行分割,然后在高分辨率遥感影像上进行灰度共生矩阵分析,选取最佳纹理参量,最后结合光谱、DEM、SLOPE信息,选择相应参数构建分类决策树,利用面向对象方法提取桉树林.最后对桉树提取结果进行精度评价,使用该方法与用Quickbird、利用TM提取精度分别为89.7%,83.1%,69.8%.表明本方法能够综合多源遥感信息,可以快速、较高精度地提取桉树人工林,具有一定的应用价值.%High resolution Quickbird images in April 2014, that of Landsat 8 during the year of 2014 and digital elevation model in 2013 in Heyuan, Guangdong province were selected as the data sources for classification of eucalyptus plantation, with decision tree and object-oriented classification method. Region of interest was selected in multidate Landsat 8, the optimal segmentation scale was calculated by estimate of scale parameter, and analysis on the high resolution remote sensing images by gray level co-occurrence matrix, best texture parameters were selected. Classification decision tree was selected by parameters from spectrum, DEM and SLOPE information, and eucalyptus plantation was classified by object-oriented method. The classification accuracy reached 89.7%.
展开▼
