Research On CME Image Registration Based On SIFT Feature Point Matching

Coronal Mass Ejection(CME)is one of the most common and violent phenomena in the solar atmosphere.It has a great impact on the space environment and safety of the earth.People have established a series of monitoring mechanisms for the characteristics of CME activities in order to mitigate this impact,and more quickly and accurately predict the outbreak of CME.As an important monitoring device,coronagraphs are indispensable.Due to many unstable factors such as equipment jitter and atmospheric disturbance,the CME images observed by the coronagraph have certain differences in image quality.In order to facilitate research,the CME images need to be processed first,and the image matching criteria is one of the most important links.Aiming at the differences in CME/C2/ C3 images,this paper proposes to use SIFT operator to perform feature extraction and registration research on C2 and C3 images.The main research contents of this paper are as follows:1.This article first introduces the research background and significance of CME image registration,briefly describes the current status of CME monitoring and image registration at home and abroad,points out the current image quality problems in the CME research process,and targets proposed to solve these problems by registering CME images.Then it briefly introduces the relevant theoretical knowledge of CME and image registration,and briefly analyzes the advantages and disadvantages of several image feature extraction algorithms.It is proposed to use SIFT operator to extract features from CME images,and then to match feature points.2.Through simulation experiments,follow the four steps of feature extraction:construct Gaussian scale space-locate the extreme point of the space-determine the direction of the extreme point-generate feature descriptors,realize image feature point extraction based on SIFT method,and match the effect Tests were conducted to prove the stability and reliability of SIFT algorithm in CME image registration.At the same time,it also points out the shortcomings of the traditional SIFT algorithm in the registration work,here mainly around two points:mismatched feature point pairs are prone to occur and there is a high space and time complexity.In view of the two shortcomings of the traditional SIFT algorithm,this paper proposes an improved method of the algorithm:for high space complexity and high time complexity,it is proposed to use KD-tree and BBF search algorithm to improve search efficiency,for mismatched point pairs,It is proposed to use the RANSAC sampling consistency algorithm to filter the points to be matched.Through simulation experiments,the improved experimental results and traditional algorithm experimental results are quantified and compared,and the effectiveness and advancedness of the proposed algorithm are obtained.In this paper,through a large number of simulation experiments,on the one hand,it proves the invariance of the SIFT operator's scale and rotation in image registration.On the other hand,it also shows that the improved SIFT-based algorithm can be well applied to CME through comparative experiments.Image registration research.