基于地貌特征的浅层崩滑体遥感自动识别

Automatic hierarchical approach to detecting shallow landslides and avalanches by the combination of multi-spectral RS imagery and DEM derivatives

建立了一种基于地貌特征的浅层崩滑体遥感自动识别方法,通过对地物的遥感光谱特性、几何形状和顺坡性相关的6次布尔运算完成对浅层崩滑体的自动识别。并以碧罗雪山一处高山峡谷地区为例,采用10m空间分辨率的SPOT-5多光谱影像和1:5万地形图生成的DEM(数字高程模型)作为数据源,进行了浅层崩滑体自动识别效果的检验。实验结果表明:①该方法顾及了地貌对浅层崩滑体空间几何形态的影响,可以有效地提高浅层崩滑体遥感自动识别的正确率;②识别对象与数据源的关系明确、决策阈值容易确定,便于使用决策树进行分类;③该方法对数据源要求较低,只需中等以上比例尺的DEM和拥有红、近红外的遥感影像即可;④粘连图斑的分割是该方法面临的主要难题。

This paper presents an automatic hierarchical approach to detecting shallow avalanches and landslides by the combination of multi-spectral RS images and DEM derivatives, which identifies these geohazards by six boolean operations based not only on their reflectance but also on geometric shapes and their geomorphic context. Corresponding experiments with 10 m resolution RS imagery and 1：50000 scale DEM in the Biluo Mountain gorge were performed to check the results of automatic recognition of the avalanche, landslide and debris flows. The results demonstrate that： 1） this approach can effectively raise the accuracy of recognition of avalanches, landslide and debris flows with the influences of their geometrical features considered; 2） it is characterized by definite relationship between the data and the recognized objects, simple logistic and easy determination of the threshold; 3） the data sources （i.e. mediumscale DEM and RS imagery with infrared and near-infrared spectra） can be widely obtained;and 4） the main problems of this method is to divide conglutinating pixels.