Research On Three-dimensional Reconstruction Of Typical Target Of Multiple Images

3D reconstruction technology has been a difficult and hot spot in the field of computer vision research.Three-dimensional reconstruction refers to the acquisition of two-dimensional image information of space objects by camera,and then the reconstruction of spatial objects' orientation,stereo shape and other sets of information by using related image processing algorithm in computer.With the development of related technology,3D reconstruction has been widely used in virtual reality,military strike and scene reconstruction.Based on the stereo vision theory,this paper studies and discusses the 3D reconstruction algorithm based on the typical targets of multiple images,and proposes an algorithm of 3D reconstruction based on multi-image,and features extraction and image matching.The key problems in camera calibration and 3D reconstruction are analyzed and studied.The main work of the thesis includes:In this paper,the theory of 3D reconstruction is introduced.By using the theory of polar geometry,the corresponding relations between multiple images of space objects are solved,and the camera parameters are calibrated by solving the fundamental matrix.In the phase of feature extraction and image matching,we improve the SIFT algorithm,introduce the custom threshold,and improve the matching accuracy of SIFT algorithm by RANSAC algorithm,and get the accurate 3D image matching points.Then dense matching of feature points is carried out by Quasi dense matching algorithm,and a dense matching point which fully reflects the information on the surface of the object is obtained.It is verified that the matching algorithm improves the stability.In the camera calibration phase,the Zhang's calibration method is introduced,and then the linear model of camera imaging geometry is formed.In view of the shortcomings of the algorithm,a nonlinear optimization algorithm is introduced into the camera self-calibration system.The parameters of the camera are calculated after the plane calibration method and LM optimization algorithm are used to complement each other.Through the verification,the improved algorithm is more convenient and the calculation process is concise.In the phase of 3D reconstruction,according to the camera calibration and stereo matching completed before,the relevant data are obtained through calculation,and the three-dimensional characteristics of the target are clearly studied.After the completion of triangulation,mapping and other work,The object is reconstructed and displayed,and the target recognition and stereoscopic reduction are realized.