Research On 3D Craniofacial Reconstruction Technology Based On Landmark

The technology of computer aided 3D facial reconstruction is widely applied in the fields of criminalistics, archeology and so on. The current status and the development of this technology are summarized, and various existing methods are analyzed and compared. Aiming at the existing problems in the technology of facial reconstruction, we put forward this research subject in order to study and improve the technology based on landmark which is used widely.Facial reconstruction based on landmark adopts the technology of surface generating via discrete points. This landmark-based method, according to the landmark of a skull and their corresponding tissue thickness, determines the landmark of facial skin and reconstructs the facial surface. Emphases and difficulties are the reconstruction of five sense organs and the determination of the tissue thickness of the landmark.The determination of five sense organs and the determination of additional landmark play an important role in degree of similarity of reproducing facial features, and the latter is the key technology. This paper presents that the reconstruction of five sense organs can be treated by dividing a face into different several subareas. In each subarea, according to its specialty, the additional landmark are conformed. Take example for the nose subarea, on the basis of BP network, the forecasting technology is given. In this subarea, the relationship between the shape of skull and the additional facial landmark can be gotten through the network training, and the station of these landmark for a skull which will be reconstructed can also be concluded. Consequently, the reproduction of a nose is realized. It has been found that, in the eye subarea, the mouth subarea, and the other subareas, the forecasting method to the determination of the additional facial landmark is simiarly available.In additional, the existing facial reconstruction methods usually apply the mean tissue thickness of landmark to calculate relevant facial landmark. These data of the thickness are statistic by experts, and each index covers a wide range of age, but to a certain skull, its age issingle and ascertainable by bone age measuring. If this skull is reconstructed using the mean index, the effect of the reproducing face is lack of individual features. Therefore, this paper advances a new method. It makes use of neural network to get the regulation that tissue thickness changes with the variance of the age. Thus, we can relatively approximate to the reasonable tissue thickness.The experimental results prove that it has better effect on 3D facial reconstruction to a certain skull. With the use of the technology developed in this paper, the quantity of data that consist of the reproducing facial surface is smaller than those of using the whole tissue thickness. In conclusion, this method is advantageous to succedent treating processes such as facial expression reproducing, facial features decorating and texture mapping and so on.