Research On 3D Facial Feature Extraction Used In Diagnosis Of Facial Morphology

As the development of 3D sensor and scanner, 3D facial image can be quickly and accurately acquired, a growing number of public 3D facial image databases and various application based on the image have become available, and 3D facial analysis technologies have been rapidly developed. 3D facial feature extraction is one of these analysis technologies, which involves simplifying the amount of resource required to describe the whole face accurately, the features are usually used for a later pattern recognition process, such as face recognition. On the background of diagnosis of children inherited mental retardation (MR) based on facial deformities, this dissertation made an extensive study of face recognition based on 3D data, and a comparative analysis of 3D facial feature extraction. A 3D facial feature extraction method is proposed based on segmentation of face based on curvature analysis. The algorithm first segments the face based on the signs of mean curvature and Gaussian curvature, then possible noses and eyes are matched in these curved regions, finally, landmark points like endocanthion, exocanthion, nasion, and pronasale are located. The cumulative 3D facial data from Zhejiang University Children Hospital was used as its input, and the description power of curvature feature has been verified by the experiment result. Because of coordinate independence, method based on curvature analysis has great potential in overcoming the problem of impact of pose and orientation in usual face detection and localization. Using the extracted feature, the dissertation utilized support vector machine to classify the faces as a diagnosis process. Finally, this dissertation designed and implemented a screening system of children inherited MR based on facial deformities via integration of the above processes. The system can diagnose inherited MR such as Down's syndrome, Fragile X Syndrome, Williams Syndrome, and DiGeorge Syndrome. The system shows that the application of computer technology to the field of genetics has improved the speed and accuracy of diagnosis.