Estimating age and synthesising growth in children and adolescents using 3D facial prototypes

https://doi.org/10.1016/j.forsciint.2018.02.024Get rights and content

Highlights

  • Describes a framework for age estimation and growth prediction from 3D photographs.

  • Accuracy of both approaches is assessed.

  • This work can facilitate person identification and building 3D facial composites.

Abstract

3D facial images are becoming increasingly common. They provide more information about facial form than their 2D counterparts and will be useful in future forensic applications. These include age estimation and predicting changes in appearance of missing persons (synthetic growth). We present a framework for both age estimation and synthetic growth of children and adolescents from 3D photographs. Age estimation accuracy was substantially better than for existing approaches (mean absolute error = 1.19 years). Our synthetically ‘grown’ images were compared to actual longitudinal images of the same cases. On average 75% of the head overall and 85% of the face were predicted correctly to within three millimetres. We find that our approach is most suitable for ageing children from late childhood into adolescence. The work can be improved in the future by modelling skin colouring and taking account of other factors that influence face shape such as BMI.

Section snippets

Cross-sectional sample

The cross-sectional sample consisted of 452 boys and 442 girls. Participants were included if they had no history of a growth disorder and if their self-reported ethnicity was Australian, European or North American. The age distribution of the sample is illustrated in Fig. 2.

Longitudinal validation sample

A subset of 24 boys and 26 girls who were imaged at age A (M = 6.92, SD = 3.72, range: 0.97–14.10) were recalled for a second image at age B (M = 12.31, SD = 4.46, range: 4.58–20.15) after an interval of several years (M = 5.40, SD = 

Age estimation

The analyses were repeated for different values of σ (30 linearly spaced values between 0.5 years and 18 years). Here we report the best estimation accuracy (when σ = 8.95 for males and σ = 11.36 for females).

The mean absolute error in age estimation for the non-linear prototyping method was 1.19 years (SD = 0.97). This was significantly lower than for the linear prototyping method (M = 1.44, SD = 1.13; paired samples t(893) = 7.76, p = <.001), which in turn was lower than the PC1 method (M = 2.24, SD = 1.72;

Discussion

In this study we developed an approach to age estimation and synthetic growth of children and adolescents using 3D facial prototyping. We have validated both approaches by: (1) comparing predicted age to chronological age and (2) comparing synthetically grown faces to actually grown faces of a subset of the sample (n = 50).

Author contributions

H.M. wrote the first draft of this manuscript, designed, implemented and performed all analysis under the supervision of P.C. This is excepting the template warping which was implemented at KU Leuven. A.P., P.C. and J.C. contributed to the initial study design and data collection and to the manuscript content and structure. All authors contributed to manuscript content and revisions. All authors have read and approved the final manuscript.

Acknowledgements

This work was supported by the Batten Foundation, the Royal Children’s Hospital Foundation, the Jigsaw Foundation, the Victorian Government Operational Infrastructure Program, an NHMRC postgraduate scholarship and a Melbourne Research Scholarship from the University of Melbourne.

We would like to thank the 3D photographers Robert Reitmaier and Lloyd Ellis. We thank the children, their families, and the schools that participated: North Melbourne Primary School, Cambridge Primary School, St Mary's

References (51)

  • V. Blanz et al.

    A morphable model for the synthesis of 3D faces

  • J. Booton

    Porn Industry's Billion Dollar New Frontier

    (2015)
  • R.G. Brereton

    The Mahalanobis distance and its relationship to principal component scores

    J. Chemom.

    (2015)
  • D.M. Burt et al.

    Perception of age in adult Caucasian male faces: computer graphic manipulation of shape and colour information

    Proc. R. Soc. Lond. B: Biol. Sci.

    (1995)
  • C. Cattaneo et al.

    Can facial proportions taken from images be of use for ageing in cases of suspected child pornography? A pilot study

    Int. J. Leg. Med.

    (2012)
  • P. Claes et al.

    Dysmorphometrics: the modelling of morphological abnormality

    Theor. Biol. Med. Model.

    (2012)
  • P. Claes et al.

    Modelling 3D facial shape from DNA

    PLOS Genet.

    (2014)
  • J. Cox-Brinkman et al.

    Three-dimensional face shape in Fabry disease

    Eur. J. Hum. Genet.

    (2007)
  • A. Dantcheva et al.

    What else does your biometric data reveal? A survey on soft biometrics

    IEEE Trans. Inf. Forensics Secur.

    (2016)
  • P. Hammond

    The use of 3D face shape modelling in dysmorphology

    Arch. Dis. Child.

    (2007)
  • P. Hammond et al.

    Fine-grained facial phenotype–genotype analysis in Wolf–Hirschhorn syndrome

    Eur. J. Hum. Genet.

    (2012)
  • P. Hammond et al.

    3D analysis of facial morphology

    Am. J. Med. Genet. Part A

    (2004)
  • T. Hassner et al.

    Example based 3D reconstruction from single 2D images

  • T. Hastie et al.

    The Elements of Statistical Learning

    (2001)
  • T. Hutton

    Dense surface models of the human face

    (2004)
  • Cited by (0)

    View full text