Research On Comprehensive Extraction Method Of Building Seismic Damage Based On Fully Polarimetric SAR

As a sudden and destructive geological disaster,earthquake seriously endangers human production activities and the safety of life and property.Among them,the collapse of buildings caused by earthquakes is one of the important causes of casualties.Therefore,the rapid acquisition and evaluation of building seismic damage information is of great significance to post-disaster emergency rescue and reconstruction.Using remote sensing technology is one of the important ways to obtain disaster information.Synthetic Aperture Radar（SAR）technology overcomes the shortcomings of traditional optical remote sensing,and can observe the earth all-weather and all-day without being affected by weather factors.In addition,SAR contains rich texture information and polarization information,which is very beneficial to describe the seismic damage of buildings.Therefore,the use of post-earthquake single-phase SAR data for building seismic damage assessment has gradually developed into a mainstream method,and domestic and foreign researchers have also made in-depth research on this.However,most of the current studies have a common problem,that is,the timeliness of seismic damage extraction from multi-temporal SAR data,and the singleness of the method in the process of information extraction.Therefore,exploring the application of multi-texture features of SAR data and improving the efficiency and universality of the new polarization feature combination in building seismic damage information extraction have become an important direction to promote the development of SAR.In order to solve the above problems,this paper is also to meet the requirements of earthquake emergency rescue loss assessment,based on the post-earthquake single Radarsat-2 fully polarimetric SAR image,this paper comprehensively uses its texture features and polarization features to extract the building areas in the disaster area and quantitatively evaluate the seismic damage of buildings.The specific research contents are as follows:1)Firstly,the methods of extracting building damage information from SAR data at home and abroad are summarized.The imaging mechanism of radar and the main theoretical basis of fully polarimetric SAR are introduced in detail.The seismic damage characteristics and classification of buildings on SAR images are analyzed.2)Then,the general situation and data of the study area are briefly introduced, and the post-earthquake single full-polarization SAR image is preprocessed by geocoding,filtering and registration with optical remote sensing image.By analyzing the limitations of various existing methods of extracting building areas using SAR texture features,and considering the feature information of local and different directional scales of the image.Therefore,a new combination of texture features is attempted,and a SAR image building area extraction method combining ULBP and Gabor texture features is proposed. The results show that this method not only improves the timeliness of calculation,but also has a more complete and more accurate extraction effect on the building area.3)Next is the extraction of building seismic damage based on the polarization characteristics of fully polarized RADARSAT-2 images.Considering the unique scattering mechanism of buildings with different damage degrees on SAR images and the applicability of various polarization characteristics of SAR images.In this paper,a multi-feature combination algorithm based on Touzi polarization decomposition and Yamaguchi polarization decomposition and circular polarization correlation coefficient is proposed to analyze the feature performance of damaged buildings on multiple feature combination images.The feature combination method with significant separability (P_dα_s1-τ₂ polarization parameter method)is selected to effectively extract the seismic damage information of buildings in the main urban area of Jiegu Town,Yushu County.4)The quantitative evaluation of seismic damage grade and seismic damage index of buildings extracted by SAR is carried out on the scale of blocks.The results show that the overall classification accuracy of this method is 88%, and the Kappa value is 0.81.Finally,the uncertainty analysis of the extraction results of building seismic damage is carried out by establishing random independent repeated tests.Specifically,on the basis of the above-mentioned extraction of building seismic damage,the blocks are taken as units,and the blocks are randomly selected to form the training and classification units of the quantitative evaluation model of seismic damage and the accuracy verification unit.Based on the classification of earthquake damage,the classification accuracy is calculated.Through multiple random independent repeated tests,the distribution curve of the classification accuracy of the extraction method is determined,and the uncertainty of the classification accuracy of the final combination method is quantitatively evaluated,so as to evaluate the seismic damage extraction ability of this research method.