Research On SAR And Optical Image Registration Method Based On Improved SIFT

With the continuous launching of remote sensing satellites in China,multisource remote sensing image sources are growing rapidly.How to combine the multi-source remote sensing images has become a hot topic in the current research.This article focuses on the registration of SAR and optical images.The research focus in multi-source remote sensing images is the registration of SAR and visible images.The optical image has strong recognition ability for different features,and can accurately reflect the gray and brightness information of the ground object,but it is always affected by the weather.However,the SAR image is an active synthetic aperture radar imaging,which can achieve all-weather imaging,and can invert many unobtrusive features.Therefore,the registration of SAR and optical images to achieve the complementary advantages of multisource remote sensing images has important research value for subsequent research work such as fusion,information extraction and object recognition.The main work is summarized as the following two aspects:(1)For the problem of difficulty in feature point extraction and slow registration speed in SAR and visible light image registration.This paper proposes a registration algorithm based on Non-Subsampled Shearlet Transform(NSST)and improved Scale Invariant Feature Transform(SIFT)algorithm.In order to improve the speed of algorithm,the algorithm uses the NSST transform to extract the low-frequency information of the image for the subsequent feature points detection.Traditional SIFT algorithm has poor robustness and few feature points in SAR image feature extraction,so the algorithm combines Harris operator and ROEWA edge detection algorithm to extract feature points in SAR image.(2)Aiming at the problem of low matching rate and matching efficiency in SAR and visible image registration,this paper proposes an improved SIFT registration algorithm based on multi-region description.In this paper,the feature description of feature points is improved to improve the uniqueness of feature points,thus improving the matching rate and matching efficiency of images.In this paper,we use GF-1 optical images and GF-3 SAR images to demonstrate that the proposed method has good performance by compared with the existing algorithms.The experimental results show that the proposed algorithm has significantly improved the number of feature points and the accuracy of registration compared with the traditional algorithms.