中文摘要：

我国1：100万的数字高程模型，是在1：5万及1：10万基本地形图上，高精度采集方里网交点高程所构建的1km分辨率地面高程数字矩阵。本文利用该DEM数据及其所派生的多种地貌信息进行地貌形态类型自动划分的技术方法。实验提取地形起伏度、地表切割度、地表粗糙度、高程变异系数、平均坡度、平均高程6个地形因子，并将各因子置于不同的信息层面中，通过主成分分析，ISODATA非监督分类法与Bayesian最大似然监督分类法相结合，对中国地貌的基本形态进行了多维信息综合分类。研究结果表明：①我国1：100万比例尺DEM在宏观地貌分类方面具有重要的价值和应用潜力；②所提取的地形因子能宏观地反映我国地形的起伏特征。为地貌形态分类提供重要的依据；③采用ISODATA非监督分类法与Bayesian最大似然监督分类法，能有效地实现我国地貌基本形态类型的定量化、自动化划分；④依据数据的统计特征进行分类，较合理地解决了类型模糊的形态宴体的归类问题。实验结果不仅揭示了此项技术在地貌形态分类中的巨大潜力，同时对于完善DEM数字地形分析的理论与方法也具有重要的意义。

英文摘要：

An automated approach was developed in terrain classification using the global DEMs with 1×1km^2 grid cells but a high sampling accuracy for their elevation data based on large-scale topographic maps. For years, the practicality and explicability of the DEMs are perplexing all the while in GIS applications. In this research, a methodology was implemented as a two-step process. Firstly, six global relief variables were derived, i.e., relief amplitude, surface incision, elevation variance coefficient, surface roughness, mean slope and mean elevation. Secondly, ISODATA unsupervised classification and the Bayesian technique of Maximum Likelihood supervised classification were applied to generate seven basic relief forms of China. The experiment revealed that 1：1,000,000 DEM of China was of practical significance and potential in macroscopic landform classification; the derived terrain variables of China contained sufficient physiographic information which provides essential basis for the automatic terrain classification. The approach permitted a quick and automatic estimation of the spatial distribution of morphologically homogeneous terrain units compared to the traditional classification methodologies. Finally, the methodology showed more objective than general DEM classification techniques owing to the mathematic methodology used and abundant quantitative relief data obtained in the study. Apparently this technique represents a splendid significance and applied foreground not only in the classification of relief forms but also in consummating the theory and methodology of DEMs based digital terrain analysis.