Image Matching Algorithm Based On Multiple Feature Fusion And Its Application

Image matching is a very important task in digital image processing.Its goal is to search for same points or areasbetween two or more images by using the effective matching algorithm.For two images of the same scene taken at different times,people are usually more concerned with the changes in these two.These changes can be divided into two types according to their causes,Natural changes and Man-made changes(Natural changes include leaves swaying and weather changes,Man-made changes include pedestrians,vehicles,etc.).The purpose of this paper is to use the image matching technology to distinguish two kinds of changes.The image matching algorithm can be divided into two basic types: gray-information-based matching and feature-based matching.The gray-information-based matching mainly uses the original gray information of the image to match,but it is sensitive to the illumination change,noise and non-rigid deformation.The problem with this type of image matching algorithm is that the matching speed is slow,and it is difficult to adapt to applications which require high real-time.Feature based image matching algorithm can overcome the shortcoming of most gray-information-based image matching algorithm.Since the number of feature points are much less than the pixels’ number in image,the computational complexity of the matching process is greatly reduced.Feature extraction is very import for such type algorithms.Edges andpoint featuresare the commonly used featuresin this type algorithms.However,the density of features in the smooth region is verylow.And the more feature points are extracted,the lager computation is needed.In feature-based image matching algorithm,only the feature points of the extracted can get match,and the other pixels are not concerned.So,it is difficult to achieve the pixel-to-pixel matching usingthis type image matching algorithm.In addition,many new ideas of image matching algorithm had been put forward.Such as,optimization algorithm based image matching,depth convolution network based image matching,image structure based image matching,etc.This paper mainly studies the matching problem of natural scene images taken by the same camera at different time,in which the camera is placed in a fixed position without change in angle and focal length.Based on the analysis of the two basic types of matching algorithms,the two algorithms cannot solve the matching problem in this paper.And the fusion of multi-level and multi feature matching can overcome the deficiency of a single type matching algorithm.Therefore,this paper proposes an image matching algorithm based on multi feature fusion.First of all,the image should be preprocessed,which is mainly to reduce the negative influence on the follow-up processing.Secondly,the fusion of multi-color image differential is used to complete the work of fine matching.In the third step,the Haar feature density mapping is used to match the local small non-rigid change region in the result of the fine matching.Finally,the multi-feature fusion method is used to match the rigid or non-rigid regions of large area.The experimental results show that the proposed method can deal with the problem of image matching in complex natural scenes,and make up the deficiency of single feature matching,and extend the field of application of image matching technology.The main contributions of this paper are as follows: a multi-color spatial fusion method,Haar characteristic density map and multi-feature fusion image matching method are proposed.