Research On UAV Aerial Image Algorithm Based On POS System Stitching

UAV aerial photography has been widely used in marine environmental supervise,sea prevention and control,natural calamity supervise and other fields with its advantages of velocity and sensitivity,low cost and easy operation.However,and it is difficult for a single image to fully display the target area.Therefore,it has become a hot topic of research in recent years to quickly assemble aerial videos or a large number of UAV sequence images to obtain panoramic images of a wide scene.In pape,geometric rectification,image enhancement,image smoothing and other preprocessing methods of aerial image are tested.Harris,SIFT(Scale Invariant Feature Transform),SURF(Speed Up Robust Features)and other feature extraction algorithms were experimented and analyzed.Finally,SURF Features were selected for feature extraction of image frames.Mismatches were deleted and image mosaic was completed by combining with the improved RANSAC algorithm(Random Sample Consensus).Aiming at the characteristics of rare feature points and single pattern of UAV aerial images in marine supervise.Article,a method on account of the data of UAVs POS(Position and Orientation System)system is raised.For the POS data in the process and circumstance,exist bound errors resulting in low data precision.Firstly,this paper built the correction equation to adjust the POS data,and used the optimied data to compute the spatial coordinates between the images and carry out registration and mosaic.The experimental results of mosaic not only carry geographic information,but also solve the problem of registration error accumulation in feature based mosaic.In order to obtain a all right,even and spontaneous panoramic image,this paper puts forwad a fusion algorithm that combines the optimum stitching line and poisson fusion,and compares it with the traditional weighted average fusion algorithm and the laplacian pyramid fusion algorithm.The appraise indicator of fusion was improved,which can effectively delete the ghosting and stitching trace problems in some areas of the image after mosaic.To sum up,the image mosaic method an account of the UAVs POS put forward in this paper,which can quickly stitch multiple aerial images and obtain a smooth and seamless panoramic image with good overall visual effects,which can effectively solve the image mosaic problem of rare and difficult to extract target feature points in marine monitoring.