This paper focuses on classifying China landform patterns automatically based on GTOPO30 data utilizing object-oriented method. In this research, six topographical factors are extracted, including: local relief,surface roughness,coefficient of elevation variation, slope variability, hillshading and mean elevation, to be combined as a "multi-spectrum image". A landform map of 1:4,000,000 of China and neighboring regions is used as the auxiliary data in the work. The result shows that the object-oriented classification of remote sensing data has advantage of utilizing the spatial information than traditional pixel based methods. The classification procedure is in line with the thinking style of human being, and the landform pattern is more complete. It is significant for increasing the accuracy and automation level for landform classification.%以GTOPO30数据为基础,采用面向对象的分类方法,进行我国地貌形态的自动划分.提取了地形起伏度、地表粗糙度、高程变异系数、坡度变化率、光照晕渲图及平均高程6个地形因子组合成特征影像,并结合《中国及毗邻地区1:400万地貌图》进行分类.研究结果表明:面向对象思想的遥感分类法可克服传统的基于像元的遥感分类难以利用空间位置信息的缺陷,分类过程更符合人的思维习惯,所分地貌类型更为完整,对提高地貌分类的精度和自动化水平具有重要的意义.
展开▼
