齿脸形状匹配关系研究

CASIA OpenIR > 毕业生 > 硕士学位论文

	齿脸形状匹配关系研究
	陈锐
	2016-05
学位类型	工程硕士
英文摘要	牙齿修复是口腔医学研究中的重要课题。在现代口腔修复医学中，选择合适的牙齿齿型以满足患者的生理、心理和社会需要成为齿型选择的重要标准。选择的齿型不但要具有自然美，还要具有社会美，即齿型应与患者的审美意识、心理、性别、年龄和职业属性等相一致。在全牙的齿型选择中，上颌中切牙的选择较为重要。在临床实践中，医生普遍通过患者的脸型来判断上颌中切牙的形状、大小、位置等信息。本课题主要针对齿脸型数据匹配关系这一研究内容，以信息更为完整的三维图像为研究材料，突破传统研究中二维图像的局限，对齿脸型特征的相关性进行分析，以此指导临床的上颌中切牙修复过程。本文主要包括以下内容：（1）提出了利用三维深度图像对齿脸型进行特征提取、分析的方法。根据北京大学口腔医院提供的齿脸型三角网格数据库，通过预处理、正交投影等操作，按深度值重采样为规整数据，将齿脸型三角网格数据转化为深度图像。利用K-means方法分别将齿脸型深度图像进行无监督聚类，并通过样本的类别验证齿脸型的对应性。研究发现不同脸型之间主要区别在于轮廓边缘和脸颊处，而不同上颌中切牙齿型主要差异位于轮廓边缘处和左右两个牙脊处。通过齿脸型类别数据统计可以发现，长型脸型与长型上颌中切牙具有较强的对应性，方圆脸型与尖圆型的上颌中切牙对应性较弱。就性别而言，女性方型上颌中切牙与方型脸型的对应性程度不高，男性齿脸型的类别统一性和对应性强于女性。（2）提出了基于几何信息和轮廓信息对齿脸型进行特征提取的方法。在脸型三维模型上利用曲率信息提取脸型的对称面及轮廓线，进而定位脸型特征点；齿型三维模型利用曲率信息和三次曲线拟合提取牙脊以及轮廓线位置。为了验证该方法的准确性和鲁棒性，我们在不同类型的齿脸型数据上进行特征提取实验，从而分析了不同类型的齿脸型在曲率和轮廓等方面的差异性。（3）提出了利用典型相关分析方法对齿脸型的相关性进行统计分析的方法。根据配准后的三维齿脸型模型，基于几何信息和轮廓信息得到的局部特征对齿脸型进行典型相关分析，探究齿脸型之间的匹配关系。实验证实脸型的长度宽度和上颌中切牙数据映射到高维空间后的特征具有相关性，表明脸型的长度和宽度对于估计牙齿形状可能提供有效信息。 ; The restoration of missing teeth is an important topic in oral rehabilitation medicine. Particularly, in the modern oral rehabilitation medicine, the restoration of missing teeth has to meet the patient's physiological, psychological, and social needs, which has been the standard of tooth selection. The restored teeth should not only have natural beauty, but also society beauty. i.e.e, the restored teeth should be consistent with the patient's aesthetic consciousness, psychology, gender, age and occupation, etc. In the oral rehabilitation medicine, the restoration of maxillary central incisors is the most important topic. In clinical practice, doctors usually select maxillary central incisors, including shape, size, location and other information, according to the patient's face. This study focuses on the relationship between shapes of maxillary central incisors data and face data. Our study adopts 3D images of teeth and faces, which provide richer information than traditional 2D images. From the 3D data, we extract depth features and investigate relationships between shapes of faces and teeth, which could guide tooth restoration in clinical practice. This thesis consists of following parts: （1）We utilize 3D range images to extract features of shapes of faces and teeth. We obtain triangular mesh surface data from the Peking University School of Stomatology. After preprocessing and orthogonal projection, 3D range images are obtained by resampling and regularizing 3D depth values . We adopt K-means clustering method to cluster range images of faces and teeth respectively, in order to validate the correspondence between the data driven clustering results. We found that that face models were different in their contour and depth of cheek, while maxillary central incisors were different in their contour and dental ridges. Statistical analysis results further indicated that face data of long type and maxillary central incisors of long type were correlated with each other. We also found stronger correlation between shapes of teeth and faces in males than in females, and no significant correlation between maxillary central incisors of square type and face of square type was found for females. （2）We utilize geometric information and contour information to extract shape features of faces and teeth. We utilize curvature information to extract symmetry plane and contour line on 3D face model, and then obtain facial feature points. For 3D tooth model, we utilize curvature information and cubic curve fitting to extract teeth ridge and contour line. In order to evaluate the accuracy and reliability of our method, we extract features of different types of face models and tooth models, and then analyze the difference between different types of teeth and faces in terms of curvature and contour. （3）We adopt canonical correlation analysis to analyze the relationship between tooth and face shape models. Based on 3D models of faces and teeth that have been registered, canonical correlation analysis is utilized to investigate relationships between faces and teeth with respect to global and local features, respectively. Experimental results have demonstrated that length and width of the face model are correlated with features of maxillary central incisors in a high-dimensional space, indicating that the length and the width of faces might be helpful to predict the shape of teeth.
关键词	齿脸型深度图像几何信息特征提取相关性
语种	中文
文献类型	学位论文
条目标识符	http://ir.ia.ac.cn/handle/173211/11555
专题	毕业生_硕士学位论文
作者单位	自动化研究所
推荐引用方式 GB/T 7714	陈锐. 齿脸形状匹配关系研究[D]. 北京. 中国科学院大学,2016.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
研究生毕业论文结题报告_final_陈锐（1470KB）	学位论文		限制开放	CC BY-NC-SA