The Research On Fast Matching Algorithm Based On Image Feature

Image matching technology has been extended to many industries,and its development has been greatly improved,with extensive and practical applications in many fields,such as image mosaic and fusion,target recognition and tracking,photogrammetry remote sensing,image retrieval and so on.It provides extremely critical technical support for image processing,and it has very important research significance for in-depth exploration of image matching technology.In the early stage,image matching is mainly used for the registration of multi-band remote sensing images after geometric correction,but now,multi-category image matching can be realized.In this paper,we do the following research on the fast matching algorithm based on image features.(1)Aiming at the problems of data complexity and time-consuming in the matching process of SIFT algorithm,etc.A singular value decomposition image matching algorithm based on block strategy is designed.First of all,the traditional SIFT algorithm is used to simulate the experiment,the experimental results show that it takes a long time after the match is completed,and the complicated calculation in the experiment process leads to the low accuracy of the matching;then the singular value decomposition algorithm is used to simulate the same experiment.The results show that the singular value decomposition matching algorithm based on the domain partition strategy can effectively reduce the complexity of the work and save the experimental time.(2)Aiming at the problem of eliminating mismatched points in the process of eliminating error matching points in RANSAC algorithm,in this paper,a function fitting method is proposed to eliminate mismatched points.First,using the singular value decomposition method to carry out image matching,the experimental results show that there are more error matching points in the matching image after matching;then the RANSAC algorithm is used to eliminate the mismatched points.The results show that the recognition degree of the proposed algorithm is lower than that of the error matching point;finally,using the function fitting algorithm to eliminate the misfit points,the accuracy and speed of the matching are improved.(3)By comparing and analyzing the experimental results of singular value decomposition algorithm,SIFT algorithm and SURF algorithm,the singular value decomposition algorithm based on block strategy can effectively solve the time-consuming and long-term problems in the experiment.Combined with the function fitting algorithm,the error matching point is eliminated,and the matching efficiency is improved by 10.66%.The research content of this paper lays a solid theoretical foundation for the realization and application of image feature matching algorithm.In this paper,there are 29 graphs,2 tables and 76 references.