Three-dimensional Skull Similarity Measurement And Gender Identification Method Research

The three-dimensional model-based skull similarity evaluation method can be used to detect the similarity between different skulls,and then provide a reference for inferring similarity of appearance,provide a reliable basis for skull gender identification,and thus help to improve the effectiveness of craniofacial reconstruction and authentication.In order to improve the accuracy of craniofacial reconstruction,this thesis conducted a skull-based similarity evaluation study in the aspects of skull feature extraction and similarity measurement.At the same time,the three-dimensional skull similarity evaluation research method was applied to the skull gender identification,which provides new ideas and methods for the study of gender identification.The main research contents of this thesis include:1.Evaluation method of skull similarity based on SPCA.The sparse principal component of the three-dimensional skull high-dimensional data is obtained by SPCA,and the test set skull data is mapped into the sparse principal component space to reduce the dimensionality of the skull data.Therefore,the threedimensional skull high-dimensional data is reduced to a low-dimensional feature vector that can characterize the skull,and the feature vector is used to find the most similar skull by two different similarity measures,mean square error and normalized inner product.The SPCA-based skull similarity evaluation method was compared with the PCA-based skull similarity evaluation method.The experimental results show that the skull features extracted by SPCA are more stable,and the accuracy,interpretability,and calculation speed of the skull similarity evaluation method based on SPCA are superior to the PCA-based skull similarity evaluation method.2.Evaluation method of skull similarity based on Wasserstein distance.Using SPCA and PCA to perform feature extraction on high-dimensional 3D skull data,the feature vectors capable of characterizing skull are obtained respectively,and then the Wasserstein distance is used as the similarity metric of the skull.Through calculating the Wasserstein distance value of the skull feature vector,the similarity between the skulls is calculated and the skull number corresponding to the skull with the highest similarity is given.The evaluation method of skull similarity based on Wasserstein distance was compared with the evaluation method of skull similarity based on mean square error and the evaluation method of skull similarity based on normalized inner product.The experimental results show that the skull similarity evaluation method based on Wasserstein distance is superior to the other two methods in terms of accuracy,reducing storage space,and avoiding decimal overflow.3.Skull sex identification method based on feature set and naive Bayes.Using SPCA to perform feature extraction on the three-dimensional skull data,the regional characteristics of the skull is obtained.Based on the skull edge information,the maximum,minimum,Euclidean distance of each axis of the nose boundary point and the cranial cavity of the skull,the rectangularity and roundness of the eyes are extracted as local features of the skull.The skull region features and local features are combined to form a skull feature set,and they are used as the input of the naive Bayes classification method.The skull gender identification results are given by the naive Bayes classification method.The highest accuracy rate of this skull gender identification method reached 91.58%.