Automated Facial Anthropometry Over 3D Face Surface Textured Meshes
- Autores: Augusto Enrique Salazar Jimenez
- Directores de la Tesis: Flavio A. Prieto Ortiz (dir. tes.)
- Lectura: En la Universidad Nacional de Colombia (UNAL) ( Colombia ) en 2014
- Idioma: inglés
- Enlaces
- Tesis en acceso abierto en: repositorio.unal.edu.co
- Resumen
The automation of human face measurement means facing mayor technical and technological challenges. The use of 3D scanning technology is widely accepted in the scientific community and it offers the possibility of developing non-invasive measurement techniques. However, the selection of the points that form the basis of the measurements is a task that still requires human intervention. This work introduces digital image processing methods for automatic localization of facial features. The first goal was to examine different ways to represent 3D shapes and to evaluate whether these could be used as representative features of facial attributes, in order to locate them automatically. Based on the above, a non-rigid registration procedure was developed to estimate dense point-to-point correspondence between two surfaces. The method is able to register 3D models of faces in the presence of facial expressions. Finally, a method that uses both shape and appearance information of the surface, was designed for automatic localization of a set of facial features that are the basis for determining anthropometric ratios, which are widely used in fields such as ergonomics, forensics, surgical planning, among others
