Research On Craniofacial Partition And Reconstruction Based On Shape Subspace

Restoring the deceased’s appearance through skull has become an interdisciplinary research hotspot in the past 20 years from a specialized subject in forensic medicine.Most of the existing computer-aided methods are from the whole skull.But the difference between people lies more in the details of the facial features.In this paper,the craniofacial partition is used for researching the craniofacial reconstruction.The main contents are as follows: 1)Collecting and pre-processing craniofacial data.267 sets of lived Uygur craniofacial data were collected and preprocessed.Firstly,the skull and face are reconstructed,denoised.The missing parts are filled and the redundant parts are removed.Then all the data are normalized in the same coordinates.At last,the physiological feature points are marked.2)Partitioning the craniofacial based on fuzzy C-means clustering.The fuzzy C-means clustering is used to divide the craniofacial into four regions: eyes,nose,mouth,and face.Compared with the previous methods,the regional results contain more effective information and are more conducive for later reconstruction.3)Extracting the subspace based on least squares canonical dependency analysis.Firstly,the least squares canonical dependency analysis is used to extract the non-linear subspace of skull and face in the shape parameter space,which maximizes the dependence between skull and face.Then,the least squares support vector regression is used for regional reconstruction.Finally,the reconstruction subareas are fused.Experiments show that the proposed method is more effective than the principal component analysis method in subspace extracting.4)Designing and implementing a craniofacial regional reconstruction system.A craniofacial regional reconstruction system is designed and implemented,and it includes the main functions,such as partition craniofacial partition,craniofacial reconstruction for different partitions and regional fusion.