DEM分辨率对地形纹理特征提取的影响

Influence of DEM Resolution on the Extraction of Terrain Texture Feature

地形纹理是区分不同地貌形态的重要依据,DEM是地形纹理分析的重要数据。然而,DEM分辨率使地形纹理特征提取存在着不确定性问题。本文以具有显著地貌多样性与差异性的陕西省为例,选择6个不同地貌类型区为研究区,以25 m分辨率DEM数据作为信息源,构建了多尺度的地面坡度、光照模拟和粗糙度数据序列。在此基础上,引入空间灰度共生矩阵(GLCM)对地形表面纹理特征进行量化分析,以揭示数据分辨率对地形纹理特征提取的影响。研究表明:对于单一样区,在DEM及其3个派生数据中,原始高程数据和粗糙度数据的纹理参数特征值,对分辨率的变化较为敏感。对于不同的地貌类型区,二阶角矩和对比度这2个纹理参数具有最大的变异系数,表明它们对于区分不同地貌类型的能力最强;二阶角矩具有较大的尺度依赖性,随着分辨率的降低,其区分能力急剧降低,而对比度对于地貌的区分能力,则随着分辨率的降低而增强,并保持在一个较大的范围内。DEM数据的对比度对于不同地貌的区分能力,在所选4个参数中最为稳定,而粗糙度数据的二阶角矩区分不同地貌的能力,随着数据分辨率的变化而最不稳定。以上结果对于根据不同的研究对象选择适宜的DEM分辨率及地形纹理参数具有一定的指导意义。

Terrain texture is an important basis to distinguish different types of landform. Due to its purity in rep- resenting terrain surface topography and its derivability in terrain analysis, the analysis of DEM based terrain tex- ture has become one of the important research subject in digital terrain analysis. But little research has focused on the scale effect of DEM based terrain texture. In this paper, DEM with 25 m resolution from 6 sample areas that represent the different landform types in Shanxi Province were selected as the source data and were further resampled into different resolutions ranging from 25 m to 325 m with a fixed interval of 50 m. The experiment dataset contains the DEM and its derivatives （slope, hillshading and roughness）. A quantitative analysis was con- ducted on the textural features using gray level co-occurrence matrix （GLCM） model to discuss the variation characteristics of DEM based terrain texture with respect to the varied data resolution. Experiments show that the parameters of DEM and roughness data are the most sensitive factors with respect to the changes in data resolu- tion. When considering different types of sample areas, the variation coefficients of angular second moment （ASM） and contrast （CON） have the biggest values among the six parameters, showing that they have the stron- gest ability to distinguish different types of landforrns. In addition, ASM has relatively high scale-dependency and its distinguish ability declines dramatically with the change of data resolution （the values of standard devia- tion change from 0.032 to 0.011 and from 0.101 to 0.038）, which indicates that ASM is suitable for recognizing the detailed terrain texture. On the contrary, the ability of CON to distinguish landform revealed an increase trend from 25 m to 325 m resolution （the values of standard deviation change from 0.145 to 0.241 and from 0.325 to 0.783）, and it has relatively low scale-dependency that indicating its suitability in recognizing the wide- range terrain texture. Overall, this study is significant for a better understanding of DEM based terrain texture, and the scale effects of other texture models will be investigated in further studies.