A Global Non-rigid Registration and Reconstruction Method for Deformable Objects

Abstract

The invention belongs to the field of computer application, three-dimensional reconstruction and registration, and aims to process the deformation of large action of a three-dimensional target object. The technical scheme employed by the invention is that according to a global non-rigid registration and reconstruction method of a deformable object, n view models are assumed and are denoted by U1, U2 to Un. For any model p, Up={u[1], u[2],***, u[N]} , N is the number of points of the model p, u[i]=(x[i],y[i],z[i],1 ) is the homogeneous coordinates of a point u[i], for pair of adjacent models Up and U(p+1), for a last model Un, an adjacent model is U1, the initial mapping relation (shown in specification) among the models is found by humans or solving a corresponding relation algorithm, namely a corresponding point, then the non-rigid transformation Xp among the models is solved, Xp={X[1],X[2], ***, X[N]}, and X[i] is a 3*4 transformation matrix. The method is mainly applied to an occasion of processing the deformation of the large action of a three-dimensional target object.

Description

technical field [0001] The invention belongs to the field of computer application, three-dimensional reconstruction and registration, and in particular relates to a global non-rigid surface alignment and reconstruction method based on sparse representation. Background technique [0002] In the field of computer graphics and computer vision, dynamic 3D reconstruction is a hot research problem, which aims to recover the corresponding dynamic scene through single or multiple cameras. With the advent of commercial depth cameras, such as Microsoft's Kinect, it has become easier and cheaper to reconstruct a model and texture of a scene. Today, it is used in many fields, such as 3D printing, games and movies. [0003] In order to achieve dynamic 3D reconstruction, some research groups have designed some multi-camera systems. Li Kun et al. (Li K, Dai Q, Xu W.Markerless Shape and Motion Capture From Multiview VideoSequences.[J].IEEE Transactions on Circuits&Systems for Video Techno...

AI Technical Summary

Problems solved by technology

Dou et al. (Dou M, Taylor J, Fuchs H, et al. 3D scanning deformable objects with a singleRGBD sensor[C] / / Computer Vision and Pattern Recognition.IEEE, 2015:493-501.) designed a single Kinect A system that allows the target object to have a large deformation, but this system requires that the deformation of adjacent viewing angles cannot be large, and it has high complexity and a large amount of calculation, so it takes a long time

Images