Research Of Three Dimensional Craniofacial Point Correspondence

The technology of craniofacial reconstruction has wide applications in archaeology,forensic medicine,anthropology,criminal detection and other fields.Compared with the traditional methods of craniofacial reconstruction,the method of reconstruction based on statistics is more objective and scientific,its result is more close to the real face.Establishing the corresponding relationship of points between skull and skull,face and face is the precondition and key step in craniofacial reconstruction based on statistical method.This thesis focuses on researching the problem of point correspondence between 3D craniofacial models,the main research work and progress include:1.The reconstruction and preprocessing of the 3D craniofacial models are implemented.It mainly includes that reconstructing for 3D craniofacial models,unifying Frankfurt coordinates,defining feature points and marking them.Through 3D reconstruction,the single 3D mesh models for skull and face are reconstructed.By preprocessing,the coordinates of different models are normalized,and all the models have the same number of feature points,Qualified data are provided for subsequent point correspondence.2.An algorithm based on the feature points deformed and multi-scale constraints for 3D craniofacial point correspondence is presented.The method mainly considered the differences of the relative position of feature points between different models.Firstly,through the deformed method of Radial Basis Function based on feature points to implement non-rigid registration between models that can make an approximate overlap between them.Then,the point correspondence is established under the multi-scale constraints.The result of the experiment validated that the method effectively improved the accuracy of point correspondence.3.An algorithm based on voxel model and multi-geometry characteristics constraints for 3D craniofacial point correspondence is proposed.The method fully considered that the points differential properties has strong expression to complexity and concavity-convexity for 3D craniofacial surface.First,the voxel model for craniofacial model is established which can reduce the search range of corresponding point.Then,by calculating the differential properties and the constraints of multi-geometry characteristics of vertices,the point corresponding relation is determined.The result of the experiment validated that the method effectively improved the veracity of point corresponding result,at the same time reduced the time complexity.4.In order to verify the feasibility and accuracy of the proposed algorithms.A 3D point corresponding system is designed and implemented.Its main purpose is to build the feature points' correspondence between template model and unknown model.