Automation of the assessment of craniofacial superimposition using soft computing and computer vision
Metadatos
Mostrar el registro completo del ítemEditorial
Universidad de Granada
Departamento
Universidad de Granada. Departamento de Ciencias de la Computación e Inteligencia ArtíficialMateria
Antropología forense Reconstrucción facial (Arqueología) Morfología Cráneo Cara Visión por ordenador Identificación Inteligencia artificial Procesado de imágenes Lógica difusa Reconocimiento de formas (Informática)
Materia UDC
681.3 3304
Fecha
2017Fecha lectura
2017-07-25Referencia bibliográfica
Campomanes Álvarez, C. Automation of the assessment of craniofacial superimposition using soft computing and computer vision. Granada: Universidad de Granada, 2017. [http://hdl.handle.net/10481/47570]
Patrocinador
Tesis Univ. Granada. Programa Oficial de Doctorado en Tecnologías de la Información y la Comunicación; Beca para la Formación de Personal Universitario (referencia AP2012-4285)Resumen
Within the forensic identi cation techniques, craniofacial superimposition is one of the most
relevant skeleton-based approaches. It involves the process of overlaying a skull with one or more
photographs of missing persons and the analysis of their morphological correspondence. This
identi cation technique has a great application potentiality since nowadays the wide majority of
the people have photographs (ante-mortem material) where their faces are clearly visible. The
counterpart, the skull (post-mortem material), is a bone that hardly degrades with the e ect of
re, humidity, high or low temperatures, time lapse, etc. Three consecutive stages for the whole CFS
process have been distinguished [DCI+11]: 1) Acquisition and processing of the materials, the skull
(the 3D model in the latest techniques) and the textitante-mortem facial photographs of the possible
candidates; 2) Skull-face overlay, which focuses on achieving the best possible superimposition of
the skull and a single ante-mortem image; 3) Decision making, where the degree of support that
the skull and the face belong to the same person (positive identi cation) or not (exclusion) is
determined.
Soft computing and computer vision present certain characteristics that become it powerful
tools to automate craniofacial superimposition, reducing the time taken by the expert and obtaining
an unbiased overlay result. In particular, evolutionary techniques for 3D reconstruction of the
skull are presented in [SCD07a, SCD+07b, SCD+09, C ACD13], highly useful for the rst stage.
In addition, the automatic system proposed in [ICDS09, ICDS11, ICD12, CAIN+14] performs
the skull-face overlay projecting the skull 3D model on the facial 2D image through a direct
correspondence between cranial and facial landmarks using evolutionary algorithms and fuzzy sets.
It also models the imprecision location of the facial landmarks in the photograph. This method
has achieved promising results, however, the nal decision of the third stage is made manually by
the forensic anthropologist in view of the superimposition obtained in the previous step.
The aim of this PhD dissertation is to extend the functionality of the current automatic CFS
procedure in order to develop a more reliable and robust computer-aided system. Concerning the
second stage of the process, we have accomplished an extension of the existing automatic methods
to superimpose a skull 3D model on a facial photograph by modeling the facial soft tissue depth
between corresponding pairs of cranial and facial landmarks. This information is available in several
anthropometric studies but its imprecise nature has caused it not to be considered in automatic
skull-face overlay methods yet. Besides, a deep study for applying the most appropriate metrics
in order to obtain the best possible superimposition has been performed. In the third stage, we
propose a complete framework for a decision support system, which takes into account all the
sources of information and uncertainty involved in the process. This decision support system has
been developed using computer vision techniques and fuzzy logic, and it has been evaluated and
validated with real positive and negative cases obtaining really good performance.