Research On The Method Ofconstructing 3D Modelbased On 2D Image

How to use the limited information of two-dimensional image to construct the object model in the image has been a hot topic in the field of computer vision.With the improvement of various technical theories and the increasing precision of optical system,the advantages of automatic and rapid 3d model generation of objects are more and more obvious by using the 3d reconstruction technology based on image.The technology can realize automatic or semi-automatic reconstruction.The traditional 3d reconstruction technology based on binocular stereo vision requires two cameras to be placed in parallel at different viewpoints to take images to calculate the depth by using parallax principle.We need two cameras in the experiment,which increase the cost of the experiment.And it is difficult to place the two cameras in parallel in practice.In addition,when extracting image feature points,the use of a single algorithm will lead to the loss of some key features of the object,affecting the effect of 3d reconstruction.This thesis discusses the technology of using one camera to take images in different angle of the same object,and using the two algorithms to extract the feature points to rebuild the model.The main work includes image acquisition,camera calibration,feature point extraction and matching,generation of triangulation model,etc.The calibration of camera is used Zhang Zhengyou calibration method to calculate.The external parameters are obtained by decomposing the essential matrix.In addition,this thesis gives the formula of solving the world coordinates of object feature points by the relationship between camera imaging models.After obtaining the world coordinates of all the feature points,the Delaunay triangulation was carried out to obtain the object triangulation model.The results of this paper show that the multi-images 3d reconstruction technology of binocular visual image based on single camera is simple to operate,low cost,and the reconstructed object model is more realistic.It can be widely used in situations where the accuracy of the model is low.