Craniofacial Reconstruction Based on Geodesic Regression Model

Abstract

<p indent=0mm>Craniofacial reconstruction refers to predicting the corresponding appearance based on the characteristics of an unknown skull. In order to solve the problem of large amount of data, manual work of calibrating feature points and difficulty in defining craniofacial feature points in the process of craniofacial reconstruction, for the three-dimensional skull and skin model represented by a triangular mesh, the craniofacial model is represented by a set of geodesics which are starting from the tip of a nose, thereby craniofacial reconstruction method based on geodesic regression is proposed. First, a set of geodesics is extracted from the nose tip point of the 3D face model; then the skull and the geodesics extracted from the corresponding face model are used as training samples, and the partial least square regression statistical model is used to reconstruct geodesics corresponding to the unknown skull; Finally, the iterative closest point algorithm is used to match the geodesics generated by the geodesic regression method with the facial statistical model to reconstruct the face of the test skull. Craniofacial reconstruction experiments have been conducted on 192 sets of craniofacial data. The reconstruction results based on geodesic regression were compared with the reconstruction results of principal component analysis and partial least squares regression. The experimental results showed that our method can improve the reconstruction accuracy, reduce the reconstruction time, and achieve a good reconstruction effect.