Method Research Of Skull-face Overlay In Computer-aided Craniofacial Superimposition

Through skull and face image overlay to identify the unknown skull remains, craniofacial superimposition plays a great role in crime scene and disaster. However, this technology exists problems of complex certification process and influences by subjective attitude. Application of computer graphics to enhance efficiency and accuracy of this identity authentication method of craniofacial superimposition is urgently needed. In this paper, we do research involved in a critical step skull-face overlay in craniofacial superimposition, including feature points extraction of face image, face pose estimation,3D-2D skull projection, and registration overlay of2D skull and face. The main work is as follows:(1)Employ an improved AAM method for automatically extracting facial feature points, combining skin similarity with gray value to draw an initial shape, and precise positions of feature points are further refined by Gabor features. Experiments show it gains lower error rate and more accuracy compared with traditional method.(2)Use a method of face pose estimation based on the geometric model parameter. Automatically calculate rotation angles relative to front face image in X, Y, and Z three directions. Results show this method can accurately estimate the face rotation angles.(3)The method of3D-2D skull projection based on camera calibration is proposed. Calculate photographic parameters of face image to ensure photographic parameters consistency of3D skull projection and facial image. Compared with existing methods, this method combines the structure of human facial features and photographic parameters to obtain more precise projection of the skull.(4)The automatically registration and overlay method of2D skull and face image is proposed. Combining prominent facial features of contours and points, and taking into account the effect of soft tissue on the overlay, register the mandibular line and finally achieve overlay result of3D skull projection and face image. Experiments show that the method can achieve more accuracy and efficiency.(5)Design and implement an application platform of craniofacial superimposition based on methods presented in this paper. The system provides interactive function of3D skull and automatic extraction of facial feature points, achieves skull-face overlay result through face pose estimation and calculation of3D skull photographic parameters.This research is supported by the National Natural Science Foundation of China (F010402,60736008,61305032).