Abstract Evidence suggests that both the interaction of so-called Merkel cells and the epidermal stress distribution play an important role in the formation of fingerprint patterns during pregnancy. To model the formation of fingerprint patterns in a biologically meaningful way, these patterns must become stationary. For the creation of synthetic fingerprints, it is also highly desirable that rescaling the model parameters leads to rescaled distances between the stationary fingerprint ridges. Based on these observations, we propose a new anisotropic interaction model for the formation of fingerprint patterns. In this model, interaction forces depend not only on the distance vector between cells and model parameters but also on an underlying tensor field representing epidermal stress. This dependence leads to complex, anisotropic patterns. We demonstrate that fingerprint patterns can be modeled as stationary solutions by choosing the underlying tensor field appropriately.
Introduction
Synthetic fingerprint generation is essential for developing and testing biometric identification algorithms while avoiding the privacy concerns and costs associated with real human data. While traditional methods like Gabor filters produce visually convincing results, they often lack the underlying structural logic of real fingerprints. Previous biological models, such as the Kücken–Champod model, attempted to simulate Merkel cell movement based on mechanical stress but often failed to reach a "stationary state"—the patterns continued to shift or merge over time. This paper introduces a model that is both biologically grounded and mathematically stable, ensuring patterns remain fixed as they do in nature after the seventh month of pregnancy.
Methodology
Anisotropic Interaction and Tensor Fields
The core of the proposed framework is an agent-based particle model where $N$ particles represent Merkel cells. Unlike standard models where forces depend solely on the distance between particles, this model incorporates an underlying tensor field representing epidermal stress. Forces are calculated anisotropically: they differ depending on whether particles are aligned with the direction of smallest stress (along the ridges) or largest stress (across the ridges). To create realistic global patterns, quadratic differentials are used to generate tensor fields for common fingerprint features like cores and deltas.
Bio-inspired Force Functions
The model utilizes a system of Ordinary Differential Equations (ODEs) where interaction coefficients are modeled as damped harmonic oscillators. Instead of simple exponential attraction or repulsion, the use of sine and cosine terms reflects common assumptions in cell biology, allowing for a more nuanced interplay between forces. This mathematical structure ensures that the system can reach a steady state, characterized by parallel, equidistant lines.
Key Results
Stability and Realism
While earlier models resulted in patterns that dissolved or over-simplified over long time intervals, the new anisotropic model reaches a clear stationary state. The simulations successfully replicate complex biological features, including "cores" (the center of a whorl) and "deltas" (triangular junctions), which are qualitatively indistinguishable from those found in nature.
Parameter Rescaling and Minutiae Formation
A significant finding is the model's ability to rescale. By adjusting a single parameter ($\eta$), researchers can shrink or expand the distance between ridges without altering the overall fingerprint shape, allowing for the generation of diverse identities from a single stress map. Furthermore, the model provides a mathematical interpretation of minutiae: ridge endings and bifurcations occur at points where attraction forces along the lines of largest stress cause parallel lines to merge, creating characteristic "Y" patterns.
Conclusion and Future Work
The proposed model successfully bridges the gap between biological theory and synthetic data generation by producing stable, realistic fingerprint patterns. Future work will focus on reducing computational complexity by moving toward continuum models (using Partial Differential Equations) and exploring 3D simulations to account for the depth of the epidermis. Additionally, formal statistical validation will be required to ensure synthetic prints are indistinguishable from real ones in automated biometric systems.
Summary updated