关键词: 形态关系建模, 生理点配准, 偏最小二乘回归, 薄板样条函数

Abstract: In order to improve the modeling accuracy of morphological relationship between skull and face, a local morphological relationship modeling method based on physiological registration of vertices is proposed. According to the correspondences of feature points, craniofacial models are preliminarily registered by using Thin Plate Spline（TPS） deformation. The optimal physiological correspondences of vertices is determined by the weighted distance of Euclidean distance and local geometric features, and a classified craniofacial model database is constructed. The morphological relationship modeling is decomposed into coordinate component modeling for each facial vertex by applying Partial Least Squares Regression（PLSR）. All coordinate calculation models, with each coordinate component of facial vertex as dependent variable and the selected significantly related coordinate components of skull vertices as independent variables, constitute the local morphological relationship model. The experimental results show that this method can describe craniofacial morphological relationship more accurately and effectively.

Key words: morphological correlation modeling, physiological registration of vertices, Partial Least Squares Regression（PLSR）, Thin Plate Spline（TPS）