基于对象的高分辨率遥感图像滑坡检测方法

滑坡遥感检测对于灾害调查和制图具有重要意义,面向对象是高分辨率遥感图像信息提取的重要思路。以高植被覆盖的热带雨林地区为例,在对试验区滑坡遥感图像特征进行分析的基础上,提出了基于对象的高分辨率遥感图像滑坡检测方法。通过多期遥感图像的多尺度、多层次分割,确定了图像对象变化检测的基本单元,构建了滑坡检测多级对象层次网络;依靠图像对象特征、上下文关系和拓扑关系等,构建了适于该实验区的滑坡检测知识规则,从而将滑坡检测整个过程包含在图像对象分析之中,并最后实现了高分辨率遥感图像的滑坡检测。实例证明,基于对象的滑坡检测方法适于热带雨林地区的信息提取,这对于其它地理环境类似地区的灾害调查与制图工作具有一定的参考价值。

Object-oriented landslide detection from remote sensing imageries with high resolution

Remote sensing detection of the landslide is very important for hazard investigation and mapping,and object-oriented technology is one of these methodologies to extract thematic information from high resolution remote sensing imageries.In this paper,the study area is located in the tropical area where the earth surface is covered with high rainforest canopy,so the landslide scars can be recognized from remote sensing imageries more easily.Based on the feature analysis of landslide scars from these remote sensing imageries,the object-oriented landslide detection methodology is presented with regard to high resolution remote sensing imageries.Firstly,two-period remotely sensed imageries are segmentated into multi-scale and different levels,so the basic units of image objects for land cover change detection have been defined,and the same time,the multiple-level object hierarchy network has been constructed for landslide detection.Followed,with image object characteristics,environmental context and object topological relationship,the knowledge rule of landslide detection suiting for this experimental area is constructed,thus the whole detection procedure is contained into image object analysis and processing,as a result,the landslide detection from high resolution remote sensing imageries has been realized.The result shows that this landslide detection method based on image objects is feasible to hazard investigation and mapping in the tropical rainforest area,and this object-oriented methodology may give a reference to hazard research and investigation in other geographical environment where is similar to this area.