On The Technical Keys Of Computer-aided Photographic Superimposition

Computer-aided photographic superimposition combines conventional photographic superimposition with computer technologies, such as computer graphics, image processing, scientific computing visualization and etc, which is an application of computer technologies in personal identification in forensic anthropology. It is valuable in investigation of criminal cases and disastrous accidents. In this dissertation, several technical keys existing in it are studied including landmarks location, three-dimensional pose estimation, estimating object distance, similarity assessment of contours, etc. The research works are summarized as follows:1. An introduction is made on the technology of computer-aided photographic superimposition and its research achievements, and several technical keys involved in it are summarized.2. A key landmarks location method is proposed for facial images, in which the algorithms of GST and ASM are improved to reduce candidates, computational complications, and get better initial parameters. The accurate localization for key facial landmarks can be obtained in this algorithm.3. The combination of computer and manual adjustment is used to locating features on 3D skull, in which the algorithm of intersection of lines-planes is improved. In the phase of modeling skull, the data of digitized skull is organized by form of triangle index, and the computational efficiency can be improved.4. For three-dimensional pose estimation from single facial photo, a method is proposed based on projective geometry and facial model, through which the correspondence between facial plane and three-dimensional positions are established. Without any auxiliary equipments and prior knowledge, the poses can be calculated with several distinct landmarks.5. Aiming at estimation of photographic object distance from facial photo, the theory of anthropometry statistics in face and projective geometry are adopted to evaluate object distance directly. This method is easier to be implemented and the optimum object distance is extended.6. Much manual operation is required in traditional similarity assessment for contours of superimposition. A scale-base assessment method is proposed to determine correspondence points between curves automatically. In this method, the identification result is presented with quantitative analyses, and the subjective factors can be decreased efficiently.