我国传统地貌基本形态类型分类强调地貌单元的完整性,界线划分沿地貌实体边界而非规则统计单元,目前尚缺乏地貌实体单元的有效自动提取方法。针对这一难点,本文提出一种基于DEM的地貌实体单元数字提取方法。利用坡度分级,并搜索相邻栅格单元、计算坡度级别内相互连通栅格的面积,建立坡度、面积阈值综合判别规则进行山地平原的自动划分;利用地形倒置、水文淹没分析,将山体划分的二维判别规则扩展到实际三维地形中,并结合地形结构线提取算法进行山体界线自动提取、确定山地地貌实体单元。结果表明,该方法符合我国传统地貌分类体系,能够较好实现山地/平原的自动划分和山体界线的数字提取。

Landforms can be described and quantified into relief classes by parametrization of digital elevation model (DEM). Several algorithms are available for automated classification of landforms, and most of the current approaches are mainly based on regular statistic unit. However, these approaches are not suitable for Chinese traditional classification system of geomorphology. Under the traditional system, integrity of geographical entities is ensured, and dividing lines of landforms are coincident with boundaries of geographical entities, the statistic units of landform classification are hereby not regular. Accordingly, algorithms of automated classification of landforms in this system are more complex, and they are still lack of effective ways to realize the automated classification of landforms under the Chinese traditional classification system. Taking one with another, the auto-classification of landform is mainly enslaved to the shortage of effective methods on the geomorphic unit auto-extration. Aimed at this problem, new quantitative and automatic methods for geomorphic unit extraction were developed in this research. Through searching for conjoint raster, neighboring cells with the same slope class were joined to the same groups. Then mountains were separated from plains by the rules based on the slope and area of the group. Boundary points of mountains were distinguished by model analysis on a reversed terrain in condition of 2-dimension. Subsequently, hydrologic modeling was carried out to extract all boundary points from real 3-dimension topography. After that, basic morphometric classes were extracted by structure lines created from the boundary points. The study shows that the results of geomorphic unit digital extraction and identification reach accuracies comparable to those of handwork. The methods developed in this article fit Chinese traditional classification system of geomorphology, and could help the digital classification of landform from DEM.