Research On Fast Mosaic Method Of Aerial Video Multi View Image

Fast mosaic of aerial video multi view image is the key technology of UAV imaging reconnaissance and mapping.Through mosaic of high spatial resolution image sequence obtained by UAV aerial photography,panoramic high-resolution image with large field of view and wide field of vision can be obtained,which has high application value in military and civil fields.However,the large number of video frames,high redundancy,low mosaic accuracy and slow processing speed restrict the development of UAV imaging reconnaissance and mapping applications.Aiming at the scientific problem of slow processing speed of aerial video image Mosaic,this paper focuses on the following research work.Aiming at the problem of redundant frames in aerial video image sequence,the key frame extraction method based on navigation information is implemented.After the experiment and defect analysis,a key frame extraction method based on the fusion of navigation information and key points is proposed.This method selects the appropriate key frame by optimizing the overlap rate and the number of reliable key points between adjacent key frames.On the premise of obtaining enough matching point pairs for image registration,the overlap rate and the number of key frames are reduced as much as possible.The experimental results on Bu et al’s aerial data show that more than 90% redundant frames can be eliminated by setting a reasonable threshold for the number of matching point pairs.In addition,compared with the key frame selection method based only on navigation information,this method also overcomes the problem of missing stitching in some overlapping regions caused by insufficient key points.Aiming at the problem of large amount of calculation and slow running speed of traditional global bundling adjustment,this paper proposes an automatic grouping bundling adjustment correction method based on closed-loop.Based on the flight trajectory of UAV,this method detects the sub closed loop and makes local binding adjustment to further improve the accuracy and speed.The experimental results on Bu et al’s aerial data and the aerial data collected in this paper show that this method achieves smaller objective function value and saves more than half of the binding adjustment time.In order to further improve the robustness of image mosaic for large amount of data,this paper adopts an incremental bundling adjustment strategy based on nearest neighbor,and proposes a complete set of algorithm flow.By selecting the nearest neighbor frames for grouping,the algorithm reduces the calculation of single binding adjustment,realizes the incremental splicing process,and obtains more stable stitching effect.This paper designs and develops a fast mosaic system for aerial video multi view images.The system is based on the contents of the preceding chapters.The test results on Bu et al’s aerial data show that,compared with the commercial software named Pix4 Dmapper,the mosaic speed of the system is increased by 36.5 times on average,and good mosaic effect is achieved,which realizes the efficient and high-quality mosaic of aerial video multi view image sequence.