Research On UAV Image Fast Mosaic Based On Feature Points

In recent years,Unmanned Aerial Vehicle(UAV)images has been widely used in fields such as battlefield reconnaissance,fire rescue,three-dimensional city model construction,land resource planning,disaster early warning and so on because of its low cost,simple operation and strong flexibility.However,the field of view of a single image shot by a UAV is limited,which is insufficient to conduct macro analysis of the shooting scene.Therefore,it is necessary to study fast and stable image Mosaic technology.A complete image Mosaic system includes four steps:image acquisition,image preprocessing,image registration and image fusion.From two aspects of image registration and image fusion,this paper focuses on studying a large number of UAV image Mosaic algorithms based on point feature registration.The main contribution points are summarized as follows:(1)Aiming at the misregistration of UAV image in the case of small coincidence area,this paper proposes an improved image registration algorithm.This algorithm introduced a new filter based on the original RANSAC(Random Sample Consensus)algorithm,which could accurately identify and filter out the misregistration in RANSAC algorithm,thus improving the generalization performance of RANSAC algorithm in small coincidence regions.The experimental results show that the improved RANSAC algorithm can perfectly filter out the error registration caused by too small coincidence area of the image.(2)In view of the problem that the whole scale parameters tend to decline in the process of a large number of image registration optimization to reduce the error error,a scale correction coefficient was designed and introduced,so that the scale decline could not reduce the magnitude of the error,so the UAV pose parameter optimization returned to the direction of searching for the best spatial consistency.Experiments show that the positive coefficient can effectively restrain the change of scale error.(3)With the increase of the number of UAV images input,the time of UAV image registration optimization becomes longer and the quality decreases.In this paper,an improved UAV image pose overall optimization algorithm is proposed.The algorithm realizes the separation of the angular parameters and displacement parameters of the three-dimensional image without man-machine image,and makes it possible to distribute the two parts of the parameters which are not related to each other,and avoids the interference between them in the overall optimization.Experiments show that this method improves the quality of registration effectively,and greatly reduces the complexity of registration calculation.