Font Size: a A A

Research On Key Technologies Of UAV Aerial Image Mosaic

Posted on:2022-07-14Degree:MasterType:Thesis
Country:ChinaCandidate:S N WangFull Text:PDF
GTID:2518306494468014Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
The UAV aerial image has high resolution to the ground and less redundant information,which can effectively save money and material resources,and has been widely used in many fields.However,some traditional image matching algorithms are difficult to obtain high accuracy due to the large color difference and rich detail texture of aerial images.In addition,the stability of the UAV platform is poor,and the flight height cannot be strictly controlled.There is a certain scale transformation relationship between the obtained images with overlapping areas,and the inaccurate scale transformation will lead to poor image fusion effect.How to achieve high-precision aerial image matching and generate a Mosaic image with good visual effect is the core problem to be solved in this paper.This paper proposes to synthesize aerial image with complete target region,unified scale and high resolution through high-precision feature matching and image fusion.In the Mosaic image,the flat transition between the two images is realized,which effectively suppresses the ghosting phenomenon,and provides a good foundation for the subsequent image interpretation and analysis.The main research contents of this paper are as follows:(1)The basic principles of aerial image Mosaic are studied in detail,including image matching method,image fusion and image transformation model.Three image matching algorithms based on point features,SIFT,SURF and ORB,are studied in depth.(2)Aiming at the shortcomings of RANSAC algorithm in eliminating the mismatched points of aerial images,VFC and PROSAC algorithm were introduced into aerial image matching,and a precise matching algorithm based on VFC and PROSAC algorithm was proposed.Firstly,the rough matching relationship was established by SIFT,SURF and ORB algorithms respectively.Then,the algorithm proposed in this paper was used to eliminate the mismatched points from the rough matching points,so as to achieve high precision matching.The results of the elimination of mismatching points between RANSAC algorithm and the proposed algorithm are analyzed objectively.(3)In order to solve the problem that the image stitched by the optimal suture algorithm is easy to produce deformation in the non-overlapping region,the SPHP transformation model is introduced to eliminate the deformation in the non-overlapping region.Using the boundary parameters obtained by the transformation model,the searching range of the optimal suture line is accurately located,and then the optimal suture line is solved by the idea of dynamic programming.Experimental results show that the optimal suture algorithm after optimization can avoid deformation in the non-overlapping area and improve the real-time performance.(4)Finally,we developed an aerial image mosaic software based on Pycharm,PyQt designer and Python as design language.The interface of the software system is simple and easy to understand,and the display of image selection and stitching result is simple and clear,with good visual effect.
Keywords/Search Tags:Aerial images, Fine match, Transformation model, Optimal seam
PDF Full Text Request
Related items