Research On 3D Reconstruction Method Based On Tilt Photography Of UAV

With the characteristics of high mobility,strong security and low cost,small UAVs have gradually been widely used in digital cities,archeological research,terrain surveys,military applications and other fields.At the same time,the rapid development of UAV tilt photography technology makes it possible to obtain high-precision aerial panoramic image data in a short time.Nowadays,with the rapid development of the information age,3D models and3 D scenes have deeper requirements in practical applications.In addition to direct modeling methods,3D digital model reconstruction methods have gradually matured and have entered the commercialization process.However,traditional 3D reconstruction methods have a series of problems such as inconvenient equipment transportation and use,expensive modeling costs,and slow running of reconstruction algorithms.In order to solve the above problems,this subject combines the UAV tilt photography technology and the 3D reconstruction method of digital images,analyzing and improving the 3D reconstruction algorithm in depth,and research on3 D model reconstruction methods around outdoor objects.Utilizing the advantages of UAV tilt photography technology,such as wide viewing angle and strong operability,this paper proposes a 3D reconstruction method based on UAV tilt photography,which aims to reduce modeling costs and improve the simplicity of modeling operations,accelerating modeling speed while ensuring model accuracy.The main content of this article is as follows:1)Introducing the UAV data acquisition platform and the theoretical basis of point cloud reconstruction.Among them,the platform includes UAV,camera gimbals,and ground control stations;in addition,it also conducts basic research on point cloud reconstruction theory from the aspects of camera calibration,image control point layout,image pre-processing,multi-view image joint adjustment,and camera attitude motion recovery.2)Proposing an adaptive feature matching algorithm based on line segment geometric constraints.To consider the rich geometric features of the main research object in this paper,an adaptive feature extraction and matching algorithm based on line segment geometric constraints are proposed.This algorithm uses the similar view theory,adopts the LSD line segment detection operator,and simultaneously constrains the linear support domain vector and geometric edge length.Finally,fusing straight line segment hypothesis estimation and outlier screening elimination.The aim is to improve the efficiency of feature extraction of UAV images and the speed of matching algorithms.3)Researching on an improved 3D reconstruction algorithm based on patch point clouds.Specifically,it includes multi-view point cloud registration algorithm,patch reconstruction theory and its key steps,PMVS-based dense reconstruction algorithm and improvement.The algorithm combines normal vector features and boundary scale rotation angle constraints,and introduces a Gaussian filter Poisson algorithm based on the point cloud set registration results.Finally,the 3D dense reconstruction result with a higher degree of refinement is obtained.4)The article takes outdoor real objects as research objects,uses aerial photography to obtain aerial image sequences,and bui lds a 3D model based on a 3D reconstruction method proposed in this paper.Then the article compares and analyzes the 3D reconstruction results of the algorithm and traditional commercial modeling software,and comes to conclude that the algorithm realize the modeling process faster and meet the basic modeling accuracy in the paper.Finally,the article summarizes and prospects the full text.