Research On UAV Image Matching Algorithm

UAV photogrammetry system has the characteristics of strong real-time performance,low cost,fast response and automation.It has gradually become one of the important technical means for obtaining spatial geographic information data.Image matching is a key step in the processing and application of acquiring drone image data.The UAV image matching method is different from the general image matching method.Firstly,the image acquired by the UAV is higher in resolution and the detail texture is extremely rich.Secondly,the stability of the UAV platform is relatively poor,and the flight attitude angle is easy to change.The same ground object may have a large difference when imaging left and right images.Finally,since the flying height of the UAV cannot be strictly controlled,there will be a certain scale conversion relationship between the adjacent image overlapping areas.How to achieve high-precision UAV image matching is a new problem in the field of remote sensing,and it is also a hot topic of current research.This paper studies the image matching method of drones.The main work is as follows:(1)According to the gray-scale and feature-based types,the research status of image matching between domestic and foreign scholars is comprehensively studied,and the advantages and disadvantages of existing algorithms and the problems in the application of UAV images are analyzed.(2)The basic principles related to image matching of drones are studied in detail,including the definition of image matching,image matching method and transformation model between images.(3)Four kinds of image matching algorithms based on point features were studied in depth,and three UAV images were used to compare the feature points.(4)In view of the shortcomings of RANSAC algorithm in UAV image mismatching point culling,the VFC algorithm is introduced into UAV image matching,and a VFC algorithm based UAV image matching method is proposed.Firstly,the SIFT algorithm and SURF algorithm are used to establish the rough matching relationship respectively.Then the VFC algorithm is used to eliminate the mismatched points in the rough matching point set to achieve high-precision matching,and the mismatching point rejection result is compared with the traditional RANSAC algorithm.analysis.According to the above research,the image matching method based on point feature is most suitable for the matching of UAV images,and the commonly used image matching algorithm based on point features has the best stability for feature extraction of UAV images.The best effect is that the UAV image matching method is based on the rough matching,and the RANSAC algorithm is used to eliminate the mismatching point.However,when there are large areas of rivers,lakes and other objects in the image,the matching method based on RANSAC algorithm is adopted.Although image matching can be achieved,the accuracy of the results tends to be low,and the problem that the correct matching points are mistakenly eliminated is easy to occur.The VFC algorithm contains more discrete matching points in the coarse matching point set,whether it is compared with the SIFT algorithm or the SURF algorithm.The combination has good applicability,and can always retain more correct matching points than the RANSAC algorithm,and better solve the problem of incorrect matching points.