Research On UAV Visual Image Matching Algorithm Based On SIFT

It has become a significant research field which involves the flight control and navigation for Micro Unmanned Aerial Vehicle(MAV)by the technology of image feature detecting and matching.The MAV carries the image acquisition equipment to obtain the sequence images realize the spatial position and pose estimation has been the cut-edge knowledge and a hot area of in the application of computer vision technology to robots and unmanned systems.Therefore,there is urgent demand for a algorithm meet the specific MAV requirement.In this paper,I research the image data taken by the MAV aim to analyze the performance of SIFT and its improved algorithms.Besides,propose a simplifying SIFT algorithm applied to the platform of MAV For solving complicated geometric deformation of the captured image.The main achievements are as follows:(1)Analyze the development SIFT algorithm as proposed and the limitations.Analyze the research status of image feature extraction and matching based on SIFT and their application in robot and unmanned system,as well as the crucial role the played in autonomous navigation of MAV.(2)Summarize and research the most active topics of research in Computer Vision.Introducing the standard image data set used in the current research,and the measurement method image similarity;Choosing the SIFT algorithm as the core studying the object for the image feature and matching which meet the requirements in the specific field of MAV.(3)Profoundly study the SIFT algorithm.Introduce the concept of the image pyramid,Gaussian blur choice,the theory of scale space,the construction of the Gaussian pyramid,the detection of spatial extremum,and determining the orientation of feature.(4)Build the experimental research platform based on OpenCV,analyze the characteristics of MAV images and select the valuable experimental data,research the effectiveness and performance of the two algorithms of SIFT and SURF(5)Summarise the optimization of improved SIFT algorithm,including data dimension reduction,Gaussian parameter selection,Orientation determination.Propose the simplify improvement on SIFT.Use standard data set and MAV visual image test the performance of optimization algorithm set and UAV visual image testing simple optimization algorithm.