AUTOMATIC IDENTIFICATION OF LANDSLIDE DAMS USING STREAM LENGTH-GRADIENT INDEX

Because of the complexity and diversity of geohazards, as well as the surroundings interference, their automatic recognitions are in a situation of great difficulty and can't meet the application requirements. Finding an algorithm to extract the common characteristics of a kind of geohazard and to distinguish them from the surroundings is effective for automatic recognition of geohazards. Landslide dams of a certain volume can bring many micro-topographic features by chain processes, such as knickpoints, which is regarded as a key for their remote sensing interpretation and field investigation. From the point of fluvial geomorphology evolution, taking the knickpoints as a common feature of landslide dams, we present a automatic recognition method of landslide dams. It is based on stream length-gradient index, and can provide the computer module by GIS project designing and coding. Using DEM, remote sensing images and geological maps, we apply the method in the upper region of Kangbumaqu, in Yadong Country, Tibet, China, and obtain the following conclusions. (1) For 1:50000 DEM, the reasonable computing interval is 300m, which can reduce data error and highlight terrain differences simultaneously. (2) The three factors with the greatest transformation to fluvial geomorphology in the study area are, in order, landslide dams, lithological changes and tectonic movement. (3) The accuracy of automatic recognition of landslide dams in the study area is 85.71%, indicating the proposed method is feasible for automatic recognition of landslide dams in high mountainous areas.