Research Of Key Techniques On Statistical Craniofacial Reconstruction

3D models are recently becoming a new type of medium after sound, images and video.So3D models have a wide range of applications in games, animation, network, mechanical, entertainment and other fields. This makes3D models retrieval and compression technology to become a research hotspot in the field of computer graphics and machine learning. Retrieval of3D models in large library, and rapid access to the retrieval results through the network still exists challenges. In this dissertation, lots of exploratory research work has been done around some key techniques3D models retrieval and compression, which include multiple features fusion, improved SKPCA method for dimensionality reduction and3D models compression method based on compressed sensing.This research work is under the support of Natural Science Foundation project of China "Research on the key technology of the semantic annotation and ontology_based retrieval for3D cultural model (No.60873094)". The main contributions of this dissertation are summarized as follows:(1) Established a multi-feature fusion framework for three-dimensional model. The new feature vector can be generated by feature vectors extracted by different algorithms in this framework. Experiments show that retrieval results of the new feature vector retrieval is better than the feature vector extracted by a single algorithm.(2) The improved SKPCA method for dimensionality reduction is proposed. This method achieve an order,second-oeder and nonlinear dimensionality reduction. It can remove redundant information to the maximum extent. Experiments show that retrieval results based on the improved SKPCA is better than SPCA.(3)The first time, using sparse representation in3D model retrieval.It can complete the3D model retrieval through solving the singular equations based on inproved SKPCA. Experiments show that sparse representation method is better than the Euclidean distance in3D model retrieval.(4) This paper provides a new compression method of3D models based on compressed sensing. The new method compared with the traditional spectrum compression method,experiment shows that the new compression method is more simple than the traditional method. Our compression method is particularly suitable for large-scale data compression.