Research Of Image Matching Algorithm Based On The Feature Points

Image matching is a basic problem in computer vision and pattern recognition,and the visual similarity between images of the same scene is calculated at different sensors,different perspectives,or different times.Image matching is an indispensable step in image analysis and digital photogrammetry,and is critical for applications such as automatic navigation,machine vision,medical image analysis,and motion estimation.The current image matching methods can be divided into three categories:image transformation domain based method,image gray level based method and feature-based method.Among these methods,because the feature-based matching algorithm is to match the features(such as curvature points,synthetic descriptors,inflection points,intersecting lines,edges,etc.)of direct matching images to greatly improve the efficiency of the operation,and for the complex transformation of the image,such as geometric distortion,different resolution or different angles and other adaptability is very good.Therefore,the feature-based image matching method is the key to the study of the paper;the main work is as follows:1.In this paper,the research and development of image matching are reviewed,and the description of image matching algorithm based on point feature is realized by expounding and analyzing the classification of image matching algorithm.2.Aiming at the shortcomings of previous algorithms,such as image blurring,noise interference and rotation,which do not have sufficient robustness,a detection algorithm based on the scale interaction of the Mexican hat wavelet extraction feature point is proposed.In the scale interaction model,with the Mexican hat wavelet instead of Gabor wavelet.Through the experimental verification,the algorithm can extract the characteristic points which are consistent with the number and the relative position from the distorted image,and prove that the algorithm has strong robustness for the rotation and blur of the image.3.An image matching algorithm combining the Mexican hat wavelet and the ORB descriptor is proposed.The algorithm firstly extracts the feature points of the image by using the granular interaction of the Mexican hat wavelet,and then uses the ORB algorithm to extract the feature points.The feature description generates the descriptor,and finally matches the descriptor with Hamming distance and bidirectional matching.Compared with other algorithms,it is found that the algorithm can maintain higher matching accuracy than other algorithms,whether it is image distortion or noise interference.By comparing the efficiency of the algorithm,it also shows the real-time performance of the algorithm.