Research On Image Matching Technology Based On Improved SURF Algorithm

Image matching plays an important role in the field of image processing,and it is also a key technology of machine vision.Numerous researchers both domestics and overseas have done a lot of research and improvement on image matching algorithms.At present,image matching technology has been widely applied to remote sensing images,production lines,medical image analysis and other fields.Image matching methods can be divided into two classes: gray-scaled image matching and feature-scaled image matching.Both methods have their advantages and disadvantages: The former has high matching accuracy but consumes a large amount of computation,the latter has a small amount of calculation and strong robust characteristics for various changes,and has been widely studied and applied.This article focuses on feature-scaled image matching algorithms.The main content includes the following three aspects:(1)Introduce several related technologies of image matching methods,including image preprocessing,image matching elements,image space geometric transformation,image performance evaluation standards,classification methods of image matching.Four feature extraction methods are studied: Harris?SUSAN?Scale-invariant Feature Transform(SIFT)and Speeded Up Robust Features(SURF).The basic principle and realization processes are described,the experimental results are given,and the advantages and disadvantages of the algorithms are analyzed and evaluated.(2)In the feature matching phase,for the Random Sample Consensus(RANSAC)algorithm to calculate the large consumption,the existence of random errors and other shortcomings,proposed an improved RANSAC matching algorithm: the algorithm reduces the wrong matching points in the initial phase,and improves the way of random sampling in the original algorithm.The proposed algorithm has improved performance over the traditional RANSAC algorithm.(3)Aiming at the defect detection problems such as small element size and high density in PCB image,an improved matching method based on SURF algorithm is proposed.The method uses curve fitting method to perform matching pair preprocessing,and uses K-means clustering algorithm to eliminate the wrong matching points.Experiments show that the algorithm improves the matching efficiency and matching accuracy.