基于图像分类方法滑坡识别与特征提取——以归州老城滑坡为例

目前遥感滑坡地质灾害解译主要是通过目视解译和人机交互式解译,根据遥感图像的影纹、色调、地形、地貌等特征,辅以软件对图像进行增强处理和叠加DEM来识别滑坡灾害,但是这些解译方法并没有充分挖掘遥感图像的光谱信息。论文以三峡库区归州老城滑坡为例,利用遥感专业处理软件ENVI对灾害点遥感影像分类,探索滑坡灾害的光谱识别与特征提取。得出图像分类方法不但可以去除一些与滑坡灾害解译无关的或干扰信息,还可以通过感兴趣区选择进行监督分类将图像划分为滑坡解译所需要的简单几种类型,有助于滑坡特征提取与统计、滑坡识别和圈定滑坡边界。

Landsides Identification Based on Image Classification: A Case Study on Guizhoulaocheng landslide in the Three Gorges Reservoir Area

At present,in the landslides hazards detection,stereoscopic images interpretation and computer-human interactive interpretation are the main methods to identify landslide hazards based on the diagnostic characteristics including shape,tone,texture,terrain,geomorphology,et al.But the spectral information of remote sensing image is not made full use of in the methods.Therefore,according to the image interpretation for landslides in the Three Gorges Reservoir area,with a case study on the Guizhoulaocheng landslide,this paper analyzes the spectral identification and characteristic extraction with the help of the classification processed by ENVI.The conclusion is that image classification function can remove some useless information and can help landslide identification and characteristic extraction by classifying the images into different classes needed.