The Design And Realization Of The UAV Image Data Rapid Registration And Automatic Stitching System

UAV aerial remote system performs better in quick obtaining high-resolutionimages especially for small areas or flying difficult areas due to its own features suchas flexibility, low-cost and large-scale high-precision.Nowadays, UAV aerial remotesensing technology not only enhances the efficiency and present situation but alsoimproves the emergency support capability of surveying and mapping results. due tothe limit of altitude and the view of camera of UAV image, only a small area iscovered by singe UAV image, since that it needs to mosaic many images in onespecific task to cover all workspaces effectively. Image matching has made a greatleap in quality and quantity for matching point precision and rate after its numerousimprovement and development since it was put forward. However, UAV image hasproperties of small picture format, large overlap degree, big rotate angle, serve imagedistortion and serious noise and occlusion, algorithm which behaves well robust indistortion and noise is needed. And algorithm based on feature matching is a goodchoice which has been widely applied in this field.Based on image matching algorithm of Scale Invariant Feature Transform(SFIT),this paper research and apply for the key technologies of fast automatic splicing ofUAV image by using c++programming algorithm. The content of this paper mainlyincludes the following several parts:The first part determines the the progress of the UAV image splicingtechnology.of this paper which describes the significance and research situation bothin domestic and overseas of the UAV image splicing technology.In the second part, VS2010platform compilation completes the fast automaticsplicing of UAV image without geographic coordinates by using c++languageprogramming. Extracting the image feature points and matching the correspondingimage points by using SIFT algorithm. And then, it eliminates false matching points and completes fine matching by applying the RANSAC. Finally it accomplishes astrip image by geometric transformation model of corresponding image points.The third part tries to solve the problem such like image fusion processing colordifference, the difference of illumination, flat-fell seam by adopting the directweighted average method.The results show that SIFT algorithm can effectively extract a large number offeature points used in image matching and be applied well in UAV image stitchingwhich lack of ground control points. However, the speed of computation will affectedby excess feature points, therefore a more effective filter method is needed to bettermeet the requirements of the efficiency of the UAV remote sensing.