Research On Skull Feature Points Calibration In Computer-Aided

At present,the feature points of the 3D model are widely used in the retrieval,reconstruction and registration of the model,the related feature points calibration technology is also the focus of the current research.Human skull model as a special three-dimensional model,its surface feature calibration is a key step in craniofacial restoration,and it has gradually applied to the current criminal investigation,plastic surgery and missing population comparison industry.However,because of the complexity and specificity of the skull model,and the skull feature points need a certain biological or medical significance.Therefore,at present,most of the calibration methods of skull feature points are mainly manual.This method is highly subjective and its accuracy depends on the experience knowledge of the calibrated personnel,and its efficiency is low.In view of the problems existing in the calibration of the skull feature points,a method of computer aided calibration of the skull feature points based on the knowledge base is proposed in this paper.The main work is as follows: 1.Construction of human skull knowledge base.First of all,Parametric representation of the skull model and generate the feature descriptor;then aiming at the problem of the human skull feature points cannot be simply equated with the feature points of the general 3D model,so,generating feature point parameters for each anthropological feature point,forming a two-dimensional feature point selection area;finally the establishment of knowledge base of the human skull that contains the standard model,the skull feature descriptor and two-dimensional feature point selection.2.A method for measuring the similarity of the skull based on geometric features is proposed.Based on the results of parameterization of human skull model,we measured the similarity between different skull models from three aspects: the volume and shape of the skull model and the Euclidean distance between the surface feature points.The results showed that the most similar template cranium could be found in the skull base for the target skull,the retrieval results are verified by the improved ICP algorithm for the registration of the skull model,the target skull and the most similar skull registration error is the minimum of 0.2164 mm,proving the validity of this method.3.A new method to calibrate the feature point of human skull based on knowledge is proposed.First of all,after normalizing and aligning the calibrated skull and the template skull in the Frankfurt coordinate system,the standard feature points on the template skull are mapped onto the calibrated skull,and the approximate position of the feature point is obtained,thereby realizing the rough calibration of the feature points;the K-D tree is used to save the point cloud data of the skull model,and the k nearest neighbor point set of the rough calibration feature point is obtained by the k nearest neighbor search;Finally,the priori knowledge of the skull standard feature points in the knowledge base is selected,a two-dimensional feature point selection area is used to achieve the accurate location of the final feature point.Experimental results show,When the value of k is taken from 2.1% to 2.8% of the vertex number of the point cloud data of the skull model,the calibration effect is good,and the final error of the feature point position is 2~3 pixel.