Research On 3D Face Registration Based On Two Dimensional Feature Points

3D face models registration is a hot spot in the field of computer research,such as 3D face recognition,digital medical images,videophone,computer games,film and television production,computer vision and so on.During scanning faces by a 3D scanner,due to the complexity of the face models and the range of measurement of the equipment and other related factors,complete and accurate face model data cannot be obtained by a single scan step,so the 3D registration of the face model data is an important part of the 3D face reconstruction.Currently,some 3D face model registration methods have some problems,such as there is a certain requirement for the initial position of the face model and the effect of the face model registration with more noise is not good.To solve this problem,a registration method based on 3D model projection image SURF feature extraction for face model is proposed.The core idea is to complete the search of 3D feature point pairs by matching the SURF eigenvalues.Firstly,two face model's dimensional projection images are acquired by 3D scanning data.Next,the SURF feature is extracted and matched for the projection images of two face models and the feature point pairs are selected.Then according to the mapping feature between projection image pixels and 3D mesh,the 3D feature point pairs are found.Finally,the method of finding transformation matrix is used to complete the registration of 3D face models.Finally,a 3D face model registration system is designed and implemented according to the algorithm ideas described above.A large number of 3D face models and other models are tested and the experiment shows that the algorithm in this thesis can clearly and completely to calculate the projected image of the 3D face model.For the acquisition of two face models projection images,the detection and matching of the feature points can be completed.Through the calculation of the Euclidean distance,the accurate feature point pairs can be screened out.According to the correspondence between the two-dimensional feature points and the space points on the 3D grid,the accurate 3D feature points and 3D feature point pairs can be found,and to registration the 3D face models.The results of the 3D face models after registration are better,and the accuracy of registration is improved effectively.