Research On The Uighur Craniofacial Morphology Analysis And Its Craniofacial Restoration Method

With the rapid development of computer graphics,craniofacial modeling and morphological statistical analysis techniques provided new research methods for the extraction and recognition of craniofacial biological features.This paper,taking reconstruction of Uighur craniofacial morphology and the appearance of the unknown skull as the research goal,constructed a statistical analysis model suitable for the geometric characteristics of Uighur craniofacial morphology to reveal the essential relationship between their skull and face.The major progress of this study are as follows: 1.In order to solve the problem of scarcity of research data for Uighur craniofacial morphology,CT data of Uighur head in Xinjiang region were collected,a skull and facial model database was developed.To perform automatic registeration of 3D skulls in the case of large differences in initial poses and the skull missing larger parts,a non-rigid three-dimensional skull registration method based on boundary correspondence was proposed.First,all the boundaries of a target skull were calculated and recognized based on their lengths and the closest distances between them.Then,the pose of the target skull was adjusted according to the boundary correspondence,so that it was matched with the reference skull coarsely.At last,accurate registration of the two skulls was gradually performed from marginal to entire area using two times of Coherent Point Drift(CPD)algorithm.The experiment results showed that the method is of stronger robustness and greater registration effeciency for the situation,position,resolution,and damage of the skull compared to the automatic registration method based on integration of Iterative Closest Point(ICP)and Thin Plate Spline(TPS).2.Aimed at repairing the damaged skulls that are common in forensics and archaeology,a skull completion method based on multiple exemplars was proposed.First,random selection of three-dimensional skull samples all the exemplars within the database were registered with a standard skull model.Then,each exemplar was divided into a missing part and a remaining part according to the given damaged skull using the same registeration method.The relationship between these parts was then obtained by a regression algorithm.At last,the relationship was used to calculate the missing part of the given skull,and finally a complete skull could be obtained by merging the two parts.The results indicated that the method can repair small and large damages well while all features of the original model were retained,besides the completion accuracy was improved.3.Semi-automatic calibration of 80 feature points on the skull and face,and identification of 62 measurement points and 43 measurement items were proposed using the methods of geometric feature statistics of craniofacial morphology and skull sex identification in Uighurs.The Euclidean distance,geodesic distance calculation,as well as angle calculation method were then used to calculate the geometric measurement of straight line distance,surface distance and angle in the craniofacial model.Finally,stepwise Fisher discriminant classification methods were applied to determine gender for the complete and incomplete skulls,reaching the accuracy of about 92%.4.For the purpose of quality improvement of craniofacial reconstruction,this study proposed a new method to establish the point correspondence of skull and face.First,the TPS registration was performed on the skull and face model and the modeles were voxelized.Then the Euclidean distance was used to determine the set of nearest feature points,while the set of core points for the initial constraint was determined using the consistency of the geodesic distance;Finally,point correspondence was established by extending the set of core points.The results showed that the accuracy and speed rate increased corresponding to the craniofacial point.5.Aiming at the shortcomings of the existing statistical indicators such as the age,fatness attributes segments that are relatively broad and resulting in the lack of personality of the restoration face,craniofacial restoration method based on soft tissue thickness was proposed.On the basis of the calibrated 80 skull feature points,the corresponding soft tissue measurement items were identified on the head and the face.The tissue depths of the landmarks for all the skulls were calculated using the cylindrical normal of the feature points as the soft tissue thickness measurement direction.Support vector regression was used to construct the functional relationship between soft tissue thickness and attributes of feature points.The soft tissue thickness of the feature points was calculated using regression function according to the properties of the skull to be restored.On these basis,a thin plate spline function is used to deform the reference face model to obtain a more accurate restoration appearance.The experimental results showed that this method obtained more accurate soft tissue thickness and improved the accuracy of craniofacial reconstruction compared with the existing methods.This study proposed an improved craniofacial morphology and skull morphological restoration research technique based on the construction of database of Uighur craniofacial model and statistical analysis.This fills the gaps the morphological studies in Uighurs in our country.The research results promote the development of computer-assisted craniofacial morphology in our country and provide data support and scientific basis for archeology,forensics and anthropology.