The Method Of Landslide Extraction With High Resolution Remote Sensing Image Combining Change Detection And Object Oriented Method

Landslide is a kind of natural disaster,which is characterized by strong outburst,wide distribution and great harm.Landslides will cause heavy losses to people's transportation,industrial and agricultural production,life and property.It is of great significance for the emergency rescue and post disaster reconstruction to obtain the information of the location and size of the landslide quickly and accurately.Remote sensing technology has the advantages of fast imaging obtaining,wide coverage and low cost.Since its emergence,remote sensing has become the most effective means of disaster monitoring.In recent years,with the rapid development of remote sensing technology,the spatial resolution of remote sensing images has been continuously improved.From high-resolution remote sensing images,more types and smaller scale landslides can be identified effectively.In addition,the texture,structure,space and other information of high-resolution remote sensing images are more abundant,and the extraction accuracy of landslide can be improved by fully using these information.Therefore,in this paper,the quick extraction method of landslide from high-resolution remote sensing images is studied,and the main contents are as follows:(1)Based on the summary of landslide extraction technology,this paper focuses on the change detection method,analyzes the characteristics of the change detection method based on pixel in the extraction of landslide information,and takes the ZY-3 satellite images for the data,the change detection method was used to extract the landslide in the Jiama mining area of Tibet.The results of the landslide extraction from the Jiama mining area in Tibet show that it is simple and rapid to extract the landslide by the principal component analysis change detection method.However,the accuracy of this method is low,only 15.47%.Besides,there are some shortcomings in the method:(1)The method mainly extracts the landslide according to the change of the spectral characteristics,and some non-landslide change information is detected as the landslide;(2)the pixel is taken as the basic unit,so that the extracted landslide is not continuous,this phenomenon is particularly prominent in high-resolution remote sensing images.(2)In order to overcome the shortcoming of landslide extraction from the change detection method based on the pixel,the object-oriented landslide extraction method is used.The image is divided into objects,and the landslide extraction rules are used to extract the landslide by using the features such as image spectrum,shape and texture.Also,taking the landslide of the Tibet Jiama mining area as an example,the experiment is carried out.The results show that the object-oriented method overcomes the discretization problem of the pixel based change detection method.The extracted landslide is spatial continuous and consistent with the actual distribution,and the correct rate of extraction is 80.83%,which is obviously higher than the traditional change detection method.However,when the landslide is extracted by the object-oriented method,the image segmentation and feature extraction is carried out based on of the whole study area image,which increases the complexity of the landslide extraction and increases the data processing capacity and processing time.(3)Based on the advantages and disadvantages of the change detection method and the object-oriented method in landslide extraction,this paper proposes the object-oriented landslide extraction method based on the change detection technology.First,the change detection method is used to extract the area where the landslide is located,and then the landslide is extracted by the object-oriented method in the changing area.The results show that the correct rate of the proposed strategy is 87.32%,which is about 7% higher than the single use of object-oriented method on correct rate,and about 72% higher than the single use of change detection method using principal component analysis method.Besides,the extraction process time required is significantly less than the object oriented method,the efficiency of landslide extraction is improved significantly.(4)In order to further verify the validity and applicability of the proposed method,a group of landslides caused by the Wenchuan earthquake in 2008 is taken as an example,and takes the before and after earthquake Google Earth image as study image.This paper uses change detection method,object-oriented method and the proposed method to extract landslides respectively.The comparison of the experimental results shows that the correct rate of landslide extraction based on the proposed method is 69.22%,which is significantly higher than that of single use principal component analysis change detection method,and is even slightly higher than that of the object-oriented method.And the extracted landslides are spatial continuous,consistent with the actual spatial distribution,extraction speed is twice faster than the object-oriented method.This paper shows that the method which combines object-oriented method and change detection method can quickly and effectively extract landslide information,which provides a reference for the rapid extraction of landslide from high-resolution remote sensing image.