Study On Large Scale Landslide Extraction Algorithm Based On SAR Image

In the world,the occurrence of large-scale landslide disaster on the production and life of human beings is very great.Therefore,how to accurately and quickly detect the field situation of the landslide area is of great significance to the rescue and rescue after the disaster,and it is also very important to detect the impact scope of the landslide disaster,which is of great significance to the assessment of the degree of damage caused by the landslide.At present,SAR image has also made some achievements in landslide disaster detection,but there is still a huge development space in the study of large-scale landslide information extraction,which is also the main purpose of this paper.The work content of this paper is summarized as follows:In general,there will be a slow surface deformation before the occurrence of landslides.Based on this principle,the surface deformation rate of the area before the landslide is extracted by using the Stanford time-series difference interferometry.Firstly,the principle of Stanford time-series differential interference technique is introduced,and the two steps of high coherence point selection and phase unwinding are introduced.This paper introduces several gray scale texture feature parameters of the gray level co-occurrence matrix method,then compares and analyzes the two key parameters of window size and azimuth when calculating the texture feature parameters,and obtains the optimal window size and the appropriate azimuth,and finally obtains the most appropriate texture feature parameter map.According to the Bhattacharyya distance formula,the Bhattacharyyat distance value of each feature parameter is calculated,and then three feature parameter maps of the top three BD values are selected.The three feature parameter maps are superimposed into a composite image through image superposition to increase the effect of ground object texture and be applied to the subsequent large-scale landslide information extraction.Then,the combined images are extracted with the change detection method,and the whole image is divided into non-landslide area and suspected landslide area.In order to further improve the accuracy of landslide information extraction,this paper combines the gray-scale texture information characteristics of SAR images after the occurrence of large-scale landslides with the Stanford time-series differenceinterference method to achieve the accurate extraction of large-scale landslide information.This article selects the guizhou qinglong,meigu county,sichuan and guizhou hezhang these three places as the study area of this paper,at the same time use the 21 JingQingLong area has access to the COSMO-SkyMed original SAR image,25 TerraSAR scene meigu area-X SAR images,and 25 TerraSAR JingHeZhang area-X original SAR image extract landslide information extraction and surface deformation rate.The field verification of the landslide results in qinglong research area shows that the extraction accuracy is 67%,which proves that the method is feasible.