Research And Implementation Of Partition-Interpolation Craniofacial Reconstruction Method Based On Knowledge Base

Craniofacial Reconstruction is a process which only uses the skull information and related knowledge to reproduce human face. It is widely used in body identification in criminal investigation, facial reproduction in archaeology, and virtual plastic surgery. This paper constructed the knowledge base depending on the tissue thickness at the feature points for different skull types; selected the knowledge based on the anthropological factors; designed a method of domain decomposed interpolation based on the select knowledge to achieve the facial reconstruction, which will provide guidance for future professional system. The research includes the following:1) Generating the sample model. A large number of live skull data and face data obtained by CT technology was collected and pretreated. A 3D model for reconstructing skull and face based on CT slices was achieved using MC algorithm.2) Detecting feature points and measuring tissue thickness. With respect to the needs of the method in this paper,58 corresponding feature points on the skull and face were defined. A manual and automatic combined feature-points detection method was used to make the result both accurate and efficient. Then a fast measurement was achieved by calculating the Euclidean distance between points.3) Constructing the knowledge base and inference strategy. The sample knowledge was created by extracting the sample tissue thickness; the sample knowledge was classified by the skull type and the tissue thickness was obtained by calculating the mean within each class, which is called the reference knowledge. The knowledge and the rules in the knowledge base were expressed by BNF and corresponding generator individually. A knowledge inference strategy based on a index table was developed to select knowledge which matches different skull types. 4) Studying craniofacial reconstruction method. The facial feature points were generated upon the skull feature points and the reference knowledge (tissue thickness) using normal distance calculation. Based on these facial feature points, a method using domain decomposition, independent interpolation and integrated meshing was proposed, which guarantees both the effect and the efficiency of the reconstruction.5) Developing SKINREC craniofacial reconstruction prototype system. Applying the result in this study, a computer-aided craniofacial reconstruction prototype system was built, which realized the sample model generation, feature points detection, tissue thickness measurement, knowledge base development, knowledge selection, facial model generation and 3D display. The system has high quality of craniofacial reconstruction and is easy and clear to operate.This research is supported by National Natural Science Foundation of China(60736008).