基于地貌特征的崩塌、滑坡、泥石流遥感自动识别

建立了一种基于地貌特征的崩滑流（崩塌、滑坡、泥石流）遥感自动识别方法，通过对地物的光谱特性、几何形状和顺坡性相关的6次布尔运算完成对崩滑流的自动识别。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，进行了崩滑流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崩滑流空间几何形态的影响，可以有效地提高崩滑流遥感自动识别的正确率；②识别对象与数据源的关系明确、逻辑简单、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。

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

This paper presents an automatic hierarchical approach to detecting avalanches, landslides and debris flows 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 1:50,000 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. medium-scale 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.