Registration And Fusion Of Optical And SAR Images Based On Features In Closed Regions

The process of matching two or more images of the same scene under different imaging conditions,such as different time and viewing angle,to find the geometric transformation relationship between images is called image registration.At present,image registration technology has been widely used in many military and civilian fields.In the process of missile-borne SAR guidance and positioning,it is necessary to register the real-time imaged image and the reference image to achieve the purpose of accurately positioning the target.However,due to limitations of the existing technical conditions,the reference images are mostly optical images,the optical images and the SAR images are from different sensors,they are heterogeneous images.Due to the different sources of imaging and the mechanism of shooting,the corresponding regions on the heterogeneous images are quite different,which brings more difficulty to the registration work.Therefore,research on heterogeneous image registration technology with high precision and satisfying the real-time requirements in the field of precision guidance has become an urgent problem to be solved.Aiming at the above problems,this thesis adopts the closed regions as the features of SAR image and optical image,and uses the affine invariant moment to describe the closed regions.An optical and SAR image registration technique based on closed regions is proposed.The main research works of this thesis are as follows:1?For the problem that the SAR image is damaged by the speckle noise,which affects the registration result,this thesis firstly studies the Frequently-used SAR image statistical filtering methods.The denoising ability of the non-local mean algorithm is efficient.The Structure Similarity(SSIM)is an evaluation index of image quality.It uses the brightness,contrast and structural information of the image,which can describe the characteristics of the image well.This thesis introduces the structural similarity index into the non-local mean filtering algorithm.A non-local mean image filtering method based on SSIM is proposed.The simulation results show that the denoising performance of this method is efficient.2?Aiming at the problem of heterogeneous image registration of SAR images and optical images,an optical and SAR image registration technique based on closed regions is proposed.Image segmentation is one of the key steps in the registration process to image registration methods based on image region features.The result of image segmentation is directly related to the accuracy of the final registration result.In this thesis,the ICM(Iterative Conditional Mode)algorithm based on Markov random field is used to segment the image,then seven Hu invariant moments are used to describe the closed regions after image segmentation,and the regional center of gravity of each region is calculated.On the basis of closed-area registration,the RANSAC(Random Sample Consensus)is used to purify the closed areas with successful registration,and the misregistration is removed.Finally,RANSAC(random sampling consistency algorithm)is used to fit the projection transformation matrix between the reference image and the to-be-registered image based on the closed regions registration.The simulation proves the effectiveness of the algorithm.The two images are registered and aligned,so that the images of the same scene are kept consistent in space.Image fusion is performed after this.The data fusion of the image combines the information of the optical image and the SAR image to make the target information more accurate and complete,improve the clarity of the target image,and make the target easier to identify.