Research On The Registration Method Of Optical And SAR Image Based On The Salient Area

In recent years,the development of image registration has been relatively rapid.It is one of the current image processing technologies.Image registration is a very important part of recognition mode,automatic navigation,diagnostic medicine,computer vision,remote sensing image processing,etc..In the field of image processing,the registration of multi-source images,such as the registration of SAR images and optical images,infrared images and optical images,has not yet had a good solution.Since the imaging mechanisms presented by optical images and SAR images are completely different,this determines the difference in radiation and geometry between the two,which will increase the difficulty of registration between images,resulting in optical images and SAR images The registration between them becomes difficult,and the information of the optical image and the SAR image cannot be effectively merged.In the current research,most of the registration methods of optical images and SAR images are not fully automated processing of images,and control points need to be selected from them,but this cannot meet the needs of fast,real-time processing,such as disasters Large amounts of remote sensing data such as emergency response,military applications,etc.Therefore,how to quickly and effectively achieve automatic registration of optical and SAR images is a problem studied in this paper.This paper summarizes and analyzes the existing registration methods,which can be roughly divided into two cases.Regarding the research focus of registration between optical images and SAR images,the paper focuses on the two difficulties of feature extraction and feature description.The main work and contributions are as follows:(1)Review the relevant literatures on remote sensing image registration at home and abroad,and explain the current status of remote sensing image registration in detail.There are two types of image registration: gray-based registration and feature-based registration.It describes the specific implementation methods andadvantages and disadvantages of these two types,which lays the foundation for this and provides ideas for future research.(2)In this paper,the visual attention mechanism is studied.The Itti visual attention model is used to extract the significant area targets of the optical remote sensing image;the TW-Itti model is used to write the relevant code to achieve the target extraction of the significant area of ??the SAR image.Large-area surface features provide a stable reference feature for image registration.(3)The registration method combining NSCT transform and Dense SIFT is proposed,and the code is written to realize the registration process of optical image and SAR image from whole to fine.The feasibility of this method for registration between images is verified,and the possible deficiencies of this method when it is used in the registration of optical and SAR images are analyzed,and it is the content of the next step of further research.