Automatic registration of multi-source images for extraction of earthquake damage information

Abstract

In order to extract the information of earthquake damage from multi-source images,the authors explored a high-accuracy algorithm of automatic image registration.Firstly,It uses Moravec operator to extract feature points and uses random sample consensus(random sample consensus,RANSAC) algorithm to reject mismatching points.Secondly,it makes a log-polar transformation according to the feature points to obtain the rotated angles and the scaled factors of the reference image and the image to be registered.After compensating coefficients of the rotated angles and the scaled factors,it uses an algorithm of sub-pixel level and phase-correlated to calculate the offset between the images and uses the interpolation algorithm to resample the registration image.At last,the images before and after the earthquake are processed as classification targets.They are used to classify and extract the earthquake damage information based on the features.The results show that the designed algorithm in this paper is of the high robustness and registration precision,and is more suitable to the registration of multi-source remote sensing images.The information of earthquake damage can be better applied to the assessment of the earthquake damage than before.