Research On Key Techniques Of Multi-View Three-Dimensional Modeling Algorithm

Three dimensional reconstruction is one of the key techniques in computer graphics and computer vision.It is a key technology to establish the virtual reality of expressing the objective world.Compared with the 2D images,the 3D geometry model can reflect more information,more intuitive and comprehensive,it can be easily understood and accepted by people.In recent decades,the rapid development of3 D industry,three-dimensional reconstruction technology to show its broad application prospects,covering television entertainment,equipment manufacturing,aerospace,biomedical,defense and military and other fields.In this paper,a multi-view 3D reconstruction algorithm is used to obtain the camera calibration parameters,and the 3D model of the object is constructed by using the reconstruction algorithm,and the key techniques of the model are Poisson surface reconstruction,texture mapping,model segmentation and so on.The work of this paper is mainly divided into the following parts:(1)The first part is the generation of multi-view cloud model.In this paper,we use the visualization algorithm based on contour to obtain the visual shell which contains the object,and then extract the point cloud data from the point cloud data.(2)The second part is a Poisson surface reconstruction.According to the point cloud model,the Poisson equation is established,and the surface of the model is built,and the 3D surface topology of the point cloud information is reconstructed,and the shape of the object surface is reduced.(3)The third part is texture mapping.A simple texture mapping method is adopted to obtain the optimal image pixels for color projection matching.Then,a weighted interpolation method is proposed for texture segmentation based on boundary texture.(4)The fourth part is based on the Laplace and K-means clustering of deformation model segmentation.For invariant deformation of non-rigid 3D model,using Laplace operator,3D spatial grid model conversion for the standard model in high dimensional Laplace spectrum domain,through the embedded global point signature,it reduces the effect of model attitude change on the segmentation results,in an appropriate space,using k-means clustering algorithm to achieve isometry invariant of grid model consistency segmentation.