A Computer Vision Non-Contact 3D System to Improve Fingerprint Acquisition

The fingerprint is one of the most important biometrics, with many acquisition methods developed over the years. Traditional 2D acquisition techniques produce nonlinear distortions due to the forced flattening of the finger onto a 2D surface. These random elastic deformations often introduce matching errors, making 2D techniques less reliable. Inevitably non-contact 3D capturing techniques were developed in an effort to deal with these problems. In this study we present a novel non-contact single camera 3D fingerprint reconstruction system based on fringe projection and a new model for approximating the epidermal ridges. The 3D shape of the fingerprint is reconstructed from a single 2D shading image in two steps. First the original image is decomposed into structure and texture components by an advanced Meyer algorithm. The structural component is reconstructed by a classical fringe projection technique. The textural component, containing the fingerprint information, is restored using a specialized photometric algorithm we call Cylindrical Ridge Model (CRM). CRM is a photometric algorithm that takes advantage of the axial symmetry of the ridges in order to integrate the illumination equation. The two results are combined together to form the 3D fingerprint, which is then digitally unfolded onto a 2D plane for compatibility with traditional 2D impressions. This paper describes the prototype 3D imaging system developed along with the calibration procedure, the reconstruction algorithm and the unwrapping process of the resulting 3D fingerprint, necessary for the performance evaluation of the method.