The project

The project aims at using a multiple-view system to contactless capture the fingerprints found on ancient clay artworks and compute the corresponding three-dimensional model, in order to obtain a metric, view-independent and less-distorted representation of the fingerprints, with respect to single-view acquisition techniques.

The detection of latent fingerprints, in fact, can be a great aid in determining the authenticity of ancient artworks, especially for clay artifacts. Clay, in fact, is often modeled barehanded, causing a great number of fingerprints left on the surface. In many cases, however, the artworks are very valuable or fragile and their latent fingerprints cannot be acquired using classical forensic methods.

Single-view contactless acquisition techniques, however, cannot properly estimate the size of the captured area, and the obtained images can suffer from perspective distortions. For these reasons, in this project we propose a two-view acquisition system, able to capture the latent fingerprints left on a clay artwork, and compute their three-dimensional reconstruction. The method permits to obtain a less-distorted model of the fingerprint, with the advantages of representing a view-independent and metric reconstruction of the fingerprints present on the artwork.

The case study: "The sleeping nymph"

In this project, we describe the application of the proposed method on “The Sleeping Nymph”, a specific clay artwork associated by experts to the famous sculptor Antonio Canova (Italy, 1757-1822).

The method

The method consists in a contactless and low-cost two-view acquisition system and in a three-dimensional reconstruction method, designed in order to capture the areas of the clay artwork containing latent fingerprints, and to compute the corresponding metric three-dimensional models. The resulting model is less-distorted than the single-view acquisitions, and represents a view-independent, metric reconstruction of the considered region, which allows to determine the size of the latent fingerprint. Moreover, the model can be processed using unwrapping algorithms and enhancement techniques.

The algorithm

The algorithm is based on image processing techniques and processes pairs of images captured using two synchronized color CCD cameras. First, a crosscorrelation-based matching technique is used in order to determine a series of matching points in the two images. The points are then triangulated and the three-dimensional model is completed with the wrapping of the interpolated texture computed from the original pair of images.

Examples of the three-dimensional models

In the next figure, it is possible to observe two examples of a three-dimensional model, the corresponding texture mapping, the particulars of the three-dimensional partial fingerprints, and two animated models:

(a) (b) (c)
Animated 3D models:
Upper view: Hi-Res | Low-Res
Frontal view: Hi-Res | Low-Res
(d) (e) (f)
Animated 3D models:
Upper view: Hi-Res | Low-Res
Frontal view: Hi-Res | Low-Res

Metric and view-independent reconstruction



The three-dimensional models of a fingerprint area computed using the proposed method permit to obtain a view-independent, metric reconstruction of the fingerprint, and to compensate to the problems related to distortions and different camera orientations.

The fingerprint can then be analyzed using a more suitable point of view. The next figure shows an example of captured pairs of images and the corresponding three-dimensional models. It is possible to observe how the three-dimensional models help in creating less-distorted acquisitions of the fingerprints:

(a) (b) (c)
(d) (e) (f)

Moreover, we propose a clear example of the capability of the described method to obtain aligned, view-independent portions of the clay artifacts:

(a) (b) (c)
(d) (e) (f)

Future developments

The three-dimensional models obtained by the proposed method can then be used in order to perform biometric recognitions. For example, it is possible to apply matching techniques based on traditional two-dimensional images or directly on three-dimensional data. Matching techniques based on two-dimensional templates can guarantee the compatibility of the fingerprint models with fingerprint images captured using touch-based sensors or traditional forensic techniques. However, they require the estimation of an equivalent two-dimensional representation of the fingerprint from the three-dimensional model. The computation of this fingerprint representation should be performed by considering the shape of the surface in which the latent fingerprint is present. Considering flat surfaces with two-dimensional impressions, in fact, it can be sufficient to extract the texture of the three-dimensional model. However, more complex surfaces require the use of unwrapping techniques.

Related publications

Refereed papers in proceedings of international conferences and workshops


  • R. Donida Labati, A. Genovese, V. Piuri, and F. Scotti, "Two-view Contactless Fingerprint Acquisition Systems: a Case Study for Clay Artworks", in 2012 IEEE Workshop on Biometric Measurements and Systems for Security and Medical Applications (BioMS 2012), Salerno, Italy, September, 2012, pp. 1-8. ISBN: 978-1-4673-2722-0. [DOI: 10.1109/BIOMS.2012.6345775][PDF] [BibTex entry]

This activity is part of the project regarding the study of 2D and 3D contactless fingerprints. A brief description of the project, and the related publications, can be found at the home page of the Industrial, Environmental, and Biometric Systems Laboratory.

Acknowledgements

The authors wish to thank Carlo Virgilio and Stefano Grandesso from "Arte Moderna e Contemporanea", Rome, Italy, for the possibility to produce the acquisitions of the artwork, as well for the profitable discussions about the artistic and historical context needed to face effectively this specific application of the proposed method.

People


Downloads

Under construction...