期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2008
卷号：XXXVII-B8
页码：351-356
出版社：Copernicus Publications
摘要：In this work we present the first results of an analysis applied to detection of landslides features using remote sensing techniques in rock masses at the Betic Cordilleras (southern Spain). After geometric and radiometric corrections, several techniques are used to facilitate a first visual approach to landslide identification, from enhancement and filtering (laplacian and textural) of panchromatic images, to colour compositions and fusions, vegetation index (NDVI) calculus and principal component analysis of multi-spectral imagery, corresponding to different sensors (Landsat ETM, Spot 5 and Ikonos). By means a GIS analysis, we compute basic statistics of whole images and pixels corresponding to different landslides typologies (rock falls, rock slides and debris flows) and in addition Kolmogorov-Smirnov coefficient to estimate the correlation between images and movements. In general terms, original panchromatic and multi-spectral bands present better correlations than processed images (filters, NDVI and PC bands), being the spectral signature different depending on landslides typology. Rock falls appear in darker zones of images while rock slides and especially debris flows appear in clearer zones. In this way, digital classification allows identify mobilized areas by typologies, but partially mixed with other land-uses such as soils, fresh rock and alluvial materials. The employment of textural filters (variance, mean euclidean distance and GLCM entropy) that present higher values in landslides zones permit the discrimination among landslides and other land-uses. The conclusion is the need of combining digital classification and textural analysis to identify landslide features or mobilized areas
