Research On SAR Image Registration Algorithm Based On Features

In recent years,with the development of radar technology,SAR image data has been widely used in various fields,such as ship identification,coastline detection,and mapping.When on access to the SAR image target detection and classification process,image registration is an essential step.Image registration is one of the key technologies in image processing,and it has been widely used in medicine,geological monitoring,artificial intelligence and other fields.However,the traditional image registration algorithm based on optical images has limitations when it is applied to SAR image registration,and it has a high rate of false detection,long processing time,and even fail to match.In recent years,various improved registration algorithms for SAR image have emerged in an endless stream.The research on image registration for SAR images has become a hot topic at present and has very important research value.This thesis focuses on feature-based image registration methods,taking into account the imaging characteristics of SAR images,and has conducted depth research on feature detection and feature matching in SAR image registration.The main research contents of this article are as follows:First of all,because of the multiplicative noise,SAR images have been greatly disturbed.There are some accidents in the process of SAR image feature extraction,such as high false detection rate and poor matching effect.To solve this problem,combined with the gradient ratio edge detection method,this paper proposes a GR-SIFT-based SAR image feature extraction algorithm.In this algorithm,the gradient ratio edge point detection is used in image Gaussian scale space to extract local extreme points.And an adaptive feature description method based on circular polar coordinates is presented for SAR image registration to obtains a robust feature point set.Then,in view of the large number of similar regional noise texture problems in SAR images,in order to improve the accuracy of feature matching,a feature matching algorithm based on scale direction constraints and fast sample consistency is presented in this thesis.For the feature points of SAR images,the algorithm uses a combination of initial matching and secondary matching.After using the nearest neighbor ratio strategy to obtain initial matching point pairs,a new scale direction constraint is presented to obtain the secondary matching.Finally,to perform further precise matching,the fast sample consistency algorithm is used to filter the matching point pairs and calculate the conversion model parameters to obtain the final matching result.In order to verify the stability and robustness of the algorithm in this thesis,we used three pairs of SAR images obtained by Terra SAR-X at different times and locations as experimental data,to experimentally verify the algorithm we proposed and compare it with several excellent algorithms.Experimental results show that the improved algorithm presented in this paper has significantly improved the number of feature extraction and the accuracy of feature registration,and has a good registration effect.