| In recent years,with the rapid development of computer vision and other high-tech industries,small UAVs have been widely used in smart city,terrain exploration,military,archaeology and other fields due to its flexible mobility,low cost and strong security.Nowadays,UAV tilt photography technology can obtain the target scene image sequence from different perspectives.People begin to conduct in-depth discussion and research on 3D reconstruction technology based on UAV tilt image,and commercial 3D modeling software is graduall y mature in the same time.However,compared with the research on UAV image matching,the research on 3D mesh model reconstruction technology is relatively few,and in practical application,the traditional 3D mesh reconstruction technology still has many problems,such a s high cost,inconvenient operation,slow reconstruction efficiency and so on.In view of the above situation,this paper deeply studies the tilt photography technology of UAV,combined with the digital image grid reconstruction technology,optimized the traditional grid reconstruction algorithm,and carried out experimental research around the small-scale outdoor real scene.Taking advantage of the UAV’s strong operability and convenient acquisition of image data,this paper proposes a grid reconstruction method based on UAV tilt image,which aims to improve the efficiency of model reconstruction,reduce the cost and improve the maneuverability on the basis of ensuring the accuracy of grid reconstruction model as far as possible.The main contents of this paper are as follows:This paper expounds the theoretical knowledge of UAV tilt photography technology.Including UAV image data acquisition,image control point layout,camera calibration,feature extraction and matching,sparse point cloud reconstruction,dense point cloud reconstruction and other related algorithms;in addition,UAV,PTZ,image control point and other parameters are introduced.(1)A point cloud enhancement processing algorithm that combines filtering and interpolation is implemented.Due to the personnel,light,equipment,point cloud acquisition algorithm and other reasons,the inevitable error makes the point cloud exist noise,outliers,holes and other problems.This paper proposes a point cloud enhancement processing method which combines statistical filtering,voxel filtering and bicubic spline interpolation.It eliminates abnormal points and repairs holes at the same time,which improves the overall quality of the point cloud.(2)A mesh reconstruction optimization algorithm based on point cl oud enhancement is proposed.In view of the problem that the traditional Poisson mesh reconstruction algorithm is prone to errors caused by the interference of miscellaneous points in the normal direction estimation of point cloud,this paper proposes to introduce the moving least squares optimization function into the normal direction estimation process to improve the overall quality of the normal direction estimation of point cloud.(3)In this paper,a small UAV is used to obtain the target image sequence,and a small-scale outdoor real scene is used as the experimental object.The three-dimensional model of the target scene is reconstructed by using the mesh reconstruction method proposed in this paper.At the same time,the reconstruction results are compared with the traditional commercial modeling software,which verifies that the algorithm can meet the basic modeling accuracy and the modeling efficiency is faster.Finally,the paper summarizes and prospects the whole text. |
