Research On The Tilt 3D Reconstruction And Online Service Technology Of UAV

With the continuous promotion of "Real 3D China" and the popularity of consumer-grade UAVs,the tilt 3D reconstruction of UAVs is accelerating towards popularization.At present,most of the 3D reconstruction software is based on C/S architecture and local programs,and the software operation requires certain professional knowledge,which seriously affects the compatibility and universality of the system.In addition,ordinary users don’t pay much attention to weather,sunshine and other factors when using drones to obtain images,which leads to uneven image quality of drones and affects the accuracy and effect of 3D reconstruction.The above factors seriously hinder the popularization and application of UAV tilt 3D reconstruction technology in different fields,and the software development technology based on B/S architecture and image enhancement technology provide a reliable and efficient solution.Therefore,based on the research of image enhancement technology,UAV tilt 3D reconstruction technology and software development technology based on B/S architecture,this paper designs and implements an online service prototype system for UAV tilt 3D reconstruction,and verifies the effectiveness and reliability of the algorithm,scheme and system proposed in this paper through experiments.The main research work is as follows:(1)Based on the research of algorithm based on dark channel prior theory,a defogging algorithm based on dark channel prior and side window mean filtering is proposed;Aiming at underexposed or shadowed images,based on the comparison and analysis of existing algorithms,an improved algorithm based on Retinex and gamma function is proposed,and the effectiveness of the improved algorithm is verified.(2)The key algorithms in the process of 3D reconstruction are studied.Firstly,the common feature point detection and matching algorithms are analyzed,and the adaptability of the algorithms is evaluated.Secondly,aiming at the problem that most open-source libraries of 3D reconstruction provided by computer vision community can’t restore the spatial scale of the model,a coordinate transformation scheme based on computer vision and photogrammetry principle is proposed,which realizes 3D reconstruction without ground reference and with real positioning and scale.Thirdly,the algorithms of dense reconstruction and surface reconstruction are deeply studied and analyzed,and the reconstruction process is improved.Finally,the accuracy of the reconstruction results is analyzed.The results show that the positioning accuracy of the proposed 3D reconstruction scheme without ground control points meets the 1:500 scale accuracy requirement specified in the Product Specification of 3D Geographic Information Model Data.(3)The prototype system of UAV tilt 3D reconstruction online service is designed,and the UAV image online transmission,image online processing,online 3D reconstruction and3 D visualization based on B/S framework are realized,which provides practical online 3D reconstruction schemes for users in different fields and professional backgrounds.