Research On Registration And Stitching Technology Of UAV Aerial Images

With the development of technology,UAV have the advantages of time efficient,low cost and flexible use.There are more and more application in safety inspection and disaster assessment in fields of electric power.Aerial image mosaic extracts,matches and filters images captured by UAV,then carry out image registration,and then create a panoramic image including all scenes and targets of the related images.This paper introduces a complete aerial image mosaic process.Based on the traditional image mosaic method for a single aerial strip,a multiple strip mosaic method is proposed.Meanwhile,the proposed global optimization improves the alignment degree after image registration.In the process of the panorama creation,the non-overlapping areas are segmented and the mosaic images are stitched according to the registration result.Then,the multi-resolution blending performs smoothing on stitching boundaries,and finally the panorama is obtained.The content of this paper is as follows:1.Through the traditional image mosaic technology,this paper introduces the process and the specific method of image mosaic.In image mosaic,SURF algorithm is used to extract image feature points,and RANSAC algorithm is used to eliminate mismatching points.Finally,image registration and panorama creation are carried out.2.Aiming at the problem of accumulated error of traditional image registration,a registration model of multiple strips is proposed to solve the accumulated error problem.The topological structure is constructed based on the image GPS information,and the images in the dataset can be divided into several flight stripes.After the stitching of each single flight strip,the stitching between two flight strips effectively reduce the accumulated error and improve the accuracy of image registration.In the process of creating panorama,an effective polygon mosaic model is proposed,and the mosaic effect improves by overlapping area chipping.3.In order to make more aligned in the multiple strips,the objective function is minimized by global optimization to improve the alignment accuracy in advance.In the process of synthesizing panorama based on polygon mosaic model,the boundaries of images are smoothed by multi-resolution fusion,and finally a seamless panorama is obtained.The aerial image used in this paper is the image captured by UAV in the laboratory.Through the comparison and evaluation of the stitching experiments of two image datasets,it is proved that the proposed model is more effective than the traditional stitching method.