| Sex determination from skull plays a significant role in many research fields, such as forensic medicine, anthropology, facial reconstruction and so on. Targeting this task, this thesis studied the entire process of sex determination from skull, with the CT data of 125 adults from Shaanxi Province. According to the CT data, we reconstructed the 3D skull model, based on which, we annotated each skull’s feature points and extracted some feature criteria for sex prediction. Then we studied the sex determination methods for both intact skulls and non-intact skulls. The main contents of this thesis are summarized as follows:1. On the reconstructed 3D skull model, we annotated 62 feature points that could describe the shape of a skull. From these feature points, we calculated some feature criteria, including linear distance, surface distance, and angles. For those features that hard to measure directly (curvature of geison for example), they were described with the result of curve fitting. We got 43 feature criteria in total and presented the statistics of them.2. We adopted Fisher discriminant analysis, Logistic Regression (LR), and Support Vector Machine (SVM) to predict the sex of intact skulls. The comparison of the determination results were presented. Results from 5-fold cross validation showed that, LR and SVM had better prediction accuracy (92.1% and 92.5%respectively) than traditional Fisher discriminant analysis (89.9%) regarding to the sex determination of skulls from Shaanxi.3. For non-intact skulls, we proposed a region-based sex determination method. We first applied LR to each of the seven skull regions and got seven LR sex determination models and their prediction accuracies. These seven models and their prediction accuracies were then combined to form a new model for predicting the sex of non-intact skulls. The new model makes full use of the available skull features of a non-intact skull and improves the prediction accuracy. In addition, the model has better applicability and can be applied to identify the sex of many different kinds of non-intact skulls.4. We designed and implemented a prototype system named skull criteria measurement and sex determination system. We used this system to extract skulls’feature criteria and verified the sex prediction accuracies of our sex determination methods. |