Research On 3D Reconstruction Of Parts Based On Binocular Vision Multi-view

In recent years,it has become a hot issue to reconstruct the target structure based on the image sequence of the target structure or scene.Only through the camera around the center of the target to reconstruct the corresponding three-dimensional model needs a variety of theoretical and technical support to develop.With the continuous development of computer vision,pattern recognition,image processing,computer graphics and other fields,the application of 3D reconstruction technology is more and more close to our life,such as virtual reality,augmented reality,medical assistant diagnosis,robot vision navigation and so on.In order to realize the automatic recognition and selection of parts based on binocular vision,this paper proposes a method of parts 3D reconstruction based on binocular vision multi view,which has a wide range of app lications in machine vision perception detection,3D model restoration and reconstruction and other fields.In this paper,SFM(structure from motion)algorithm is used as the starting point to study the 3D reconstruction of parts based on binocular vision multi view.Because SFM algorithm is based on the target structure restoration of monocular image sequence,it is easy to be affected by the horizon error and the different brightness of the environment in the process of 3D reconstruction,which results in the poor robustness of the reconstruction results.In view of the above problems,this paper introduces the binocular vision theory in SFM algorithm.Firstly,by building a binocular camera,and adjusting the image plane of the left and right cameras to make them in the same horizontal plane,then using binocular camera calibration method to solve the internal and external parameters of the camera for the adjusted binocular camera,and then setting up the structure image sequence of the target parts captured by the binocular camera The volume correction makes the pixel points in the corresponding pair of left and right images in the row alignment state.Then it focuses on the research of 3D reconstruction algorithm based on binocular image sequence,and de scribes the algorithm flow in detail.Through the 3D reconstruction algorithm based on the binocular image sequence,the reconstructed left and right point clouds are combined to get the final 3D point cloud of the target part structure,and then the feasibility of the algorithm is verified by experiments.After obtaining a more accurate 3D point cloud model of the target part structure,the surface reconstruction algorithm and texture mapping method of the point cloud model are studied.In order to make th e reconstructed 3D point cloud model surface have good sealing property and similar surface features with the target structure to be reconstructed,this paper realizes the surface reconstruction of the target structure 3D point cloud by solving the Poisson equation through Poisson surface reconstruction,and finally carries on the texture mapping to the part 3D model to get the final target structure 3D model with texture information.