关键词: 地形纹理, DEM, 量化分析, 灰度共生矩阵

Abstract:

Terrain texture is an important natural texture. DEM based terrain texture attracts more attention in the research area for its purity in representing surface topography and its derivability in terrain analysis. In this paper, 10 sample areas from different landform types of Shaanxi Province were selected to make a quantitative analysis on the terrain texture by Gray level co-occurrence matrix (GLCM) model. Experiments show that, when using the DEM data with 25m resolution, the suitable analytic distance of GLCM model is not less than 3 pixels. Among all the parameters in the model, correlation could be used for texture direction detection. Contrast, variance, and different variance could be applied for texture periodicity analysis. Entropy, angular second moment and inverse different moment are suitable for texture complexity investigation. In this research, quantitative analysis is conducted to terrain texture by using DEM data, hillshade data, slope data and curvature data. The terrain texture directivity experiment shows that the correlation of hillshade data reacts sensitively to the terrain texture direction and can detect main terrain texture direction. The correlation of slope data reacts obviously in rugged topography such as hilly region and mountainous regions so it can play an auxiliary role for hillshade data in the detecting of terrain texture direction. Results of terrain texture periodicity and complexity analysis shows that among DEM data and its derived data, the mean variation coefficient of each texture parameter based on hillshade data is the highest, and it further proves that the hillshade data is most suitable for quantitative analysis of terrain texture. Quantification is conducted by variance of hillshade data to texture periodicity of different terrain texture, variance eigenvalue of flat, platform, hill and mountain region gradually increases which indicates the increase of terrain texture periodicity. Analysis is also conducted to the terrain texture complexity through angular second moment parameters computed by hillshade data. Eigenvalue has clear peak value in the sample region of flat and the eigenvalue of platform decreases obviously. Eigenvalue of hills and mountain region verge to zero which shows that texture of plat has lowest complexity, followed by the lower complexity of platform and the highest complexity of hills and mountain region. This paper also proposed a multi-parameter integrated method which employs both comprehensive periodicity and comprehensive complexity in terrain texture quantitative analysis. This method not only reduces replicate analyses but also makes full use of various texture parameter information, it also unifies range through normalization for the convenience of quantitative analysis. The result showed that these two parameters have significant response to the different terrain texture, which shows a great potential in landform recognition and classification.

Key words: quantization analysis, gray level co-occurrence matrix, DEM, terrain texture